Wonda CLI

Wonda CLI is a content creation toolkit for terminal-based agents. Use it to generate images, videos, music, and audio; edit and compose media; publish to social platforms; and research/automate across LinkedIn, Reddit, and X/Twitter.

Install

If wonda is not found on PATH, install it first. The recommended installs are the signed desktop installers (CLI + tray icon + always-on relay, zero extra steps): macOS brew install --cask degausai/tap/wonda-app (or the wonda-macos.pkg from releases), Windows winget/wonda-windows-setup.exe. The CLI-only channels below work everywhere and can add the desktop app later with wonda app install:

# npm
npm i -g @degausai/wonda

# Homebrew
brew tap degausai/tap && brew install wonda

Setup

• Auth: wonda auth login (opens browser, recommended) or set WONDA_API_KEY env var
• Verify: wonda auth check

OAuth connector auth

Claude web and Cowork connectors use Wonda's OAuth 2.1 flow instead of a CLI API key field. The connector signs in through Wonda in the browser, grants the requested account access, and receives OAuth tokens bound to the Wonda API resource. The server swaps those tokens to the account's internal API key only inside Wonda, so agents and connector hosts never see the sk_... key. For the CLI and local stdio MCP path, keep using wonda auth login or WONDA_API_KEY.

Claude Cowork local relay

Claude Cowork (the desktop app) runs local MCP servers on the host, so it can load the .mcpb bundle or a local stdio wonda-mcp config directly, WAB writes included (verified 2026-07-07). Claude web cannot. The Wonda local relay is the alternative path: it lets the REMOTE connector (web or Cowork) run actions on the user's own Mac and residential IP without any local MCP config:

1. Open https://wonda.sh/download while signed in and install the notarized Mac package.
2. Pair the relay with wonda relay pair or the first-run browser handoff. This uses the existing cli-auth flow with a relay-scoped wrelay_... credential stored in the macOS Keychain. Do not ask the user to paste an API key or device code.
3. Open https://wonda.sh/setup, connect LinkedIn, X, and Reddit through the headful local WAB, then approve the Wonda connector once in Claude.

The engine policy is auto | my_machine | cloud. auto uses the local relay when it is online and cloud otherwise. my_machine must not silently fall back: if the relay is offline, ask whether to switch to cloud.

Organizations & spend context

Wondercat orgs are shared wallets with their own seats and billing. Members can spend from the org wallet (instead of their personal credits) by switching context:

• wonda organizations list (aliases: wonda orgs list, wonda org list) — see every org you belong to with your role and seat plan in each.
• wonda use --org <slug> — sticky org context for this machine. Sets X-Wonda-Org on every request; holds, charges, and wonda balance route through the org wallet.
• wonda use --personal — back to personal.
• wonda usage — spend-only usage summary (total + per-model + per-project breakdown) for a period (--month 2026-05, or --from/--to; defaults to the current month, UTC). --project <name> restricts the report to one project. In org context it reports org-wide usage including a per-member breakdown — admin/owner role required. Admins can also download a full Excel report from the org page on the web.

Projects (spend tagging)

Projects attribute spend to a named workstream for monitoring. Agents should check the active project at task start (wonda use prints it) and set one per task when the operator monitors spend by project:

• wonda use --project <name> — sticky: every subsequent charge carries the project (in wonda usage, the API, and the org Excel report). wonda use --no-project stops tagging; switching org/personal context clears the project automatically (projects are per-scope).
• --project <name> on any command — one-off override for that invocation.
• wonda project list|create|delete — manage the registry in the active scope. Org projects are created by org admins/owners only; personal projects are self-service. Tagging against a name that doesn't exist fails with unknown_project (no silent new buckets, so typos can't split the monitoring data).

wonda topup always tops up your personal wallet, regardless of context. Topping up the org wallet (and configuring auto top-up) is admin-only and happens on the web at /organizations/<slug>. If a member runs out of org credits, the error tells them to ask an admin or switch back to personal — they cannot top up the org wallet from CLI.

Roles inside an org are separate from the seat plan:

• Owner: the original creator. Cannot be demoted or kicked. Can transfer ownership to another member from the org page (rare).
• Admin: can invite (single or bulk via paste), kick, change roles, change seats, top up, configure auto top-up, change monthly limits.
• User: can only spend within the org wallet (subject to a per-member monthly limit if the admin set one).

A paid org seat (WONDA / WONDA_PREMIUM) grants the same paid feature access (skills, etc.) as a personal paid plan, but only while in org context. wonda use --personal falls back to the user's personal account plan.

Access tiers

Wonda is paid-only: every product surface requires a paid plan, including API calls, local platform reads, browser automation, media composition, file editing, diagnostics, generation, publishing, scraping, analysis, skills, relay operation, and cloud twin. Before any product command does work, the CLI performs a live entitlement check against Wonda. The check fails closed when access cannot be verified. New accounts get no product access until they subscribe, and a positive credit balance does not substitute for a plan.

If a command returns a 403 (paid_plan_required), subscribe at https://wonda.sh/account.

This applies to the local stdio MCP path too because MCP executes the same CLI. Platform cookies and data remain on-device, but the command must first receive a successful paid-access decision from GET /api/v1/auth/access. A relay-only installation may use its scoped wrelay_... credential for that entitlement probe. The scoped credential still cannot call ordinary account APIs.

Voice cloning

Clone a voice from a 10s+ audio clip and use it in TTS. Hard limit: 20 cloned voices per account. Cost: $1.50 per clone.

# Clone from a local file (auto-uploads to media library first)
wonda voice create "Andu" --file ./sample.mp3 --description "My voice"

# Clone from existing wonda media
wonda voice create "Brand" --media-id <uuid>

# Optional source-audio preprocessing
wonda voice create "Clean" --file ./raw.wav --noise-reduction --normalize-volume

# List cloned voices (each row reports isExpired and expiresInDays)
wonda voice list

# One voice
wonda voice get <voice-id>

# Rename / re-describe (local only, no provider call)
wonda voice update <voice-id> --name "New Name" --description "..."

# Delete
wonda voice delete <voice-id>

Use a cloned voice in TTS by passing the providerVoiceId from voice get as voiceId to /audio/speech:

wonda audio speech "Hello world" \
  --model minimax-speech-2-8-hd \
  --params '{"voiceId":"<providerVoiceId>"}'

7-day expiry: cloned voices that haven't been used in TTS within 7 days are automatically expired. Running TTS with a cloned voice automatically refreshes its expiry. Idle voices that lapse must be re-cloned ($1.50 again).

Credentials vault

Persist logins created on external platforms (Instagram, TikTok, Twitter, etc.) so they can be reused on the next run. Passwords are AES-256-GCM encrypted with a server-side key and only decrypted on get.

# Create
wonda credentials create --website instagram.com --username myhandle \
  --email me@example.com --password-stdin <<< "hunter2" \
  --metadata '{"signup_source":"wonda-email"}'

# List (passwords omitted)
wonda credentials list --website instagram.com

# Get full record including decrypted password
wonda credentials get <id>

# Update any field (use --password-stdin to rotate; --username "" to clear)
wonda credentials update <id> --username newhandle

# Delete
wonda credentials delete <id>

# Fetch + record why you're using it in one call — POST, not GET, because
# it writes a 'used' event with the reason. Prefer this over `get` whenever
# you can articulate the reason.
wonda credentials use <id> --reason "instagram signup flow"

# See recent events (created / used / rotated / updated) for audit
wonda credentials events <id>

Fields: website (required — typed input like insta is canonicalized to instagram.com), username, email, password (required), metadata (arbitrary JSON). At least one of username / email must be present. Multiple records per (website, username) are allowed — dedupe on your side if you need to.

Event log: every credentials get/use, create, password rotate, and other updates are recorded as events on the credential (actor: cli | web | system). Use credentials events <id> or the web UI's history icon to audit. The event log is append-only and cascades on credential delete.

Global output flags

All commands support these output control flags:

• --json — Force JSON output (auto-enabled when stdout is piped)
• --quiet — Only output the primary identifier (job ID, media ID, etc.) — ideal for scripting
• -o <path> — Download output to file (implies --wait)
• --fields status,outputs — Select specific JSON fields
• --jq '.outputs[0].media.url' — Filter JSON output with a jq expression

CLI announcements & deprecation warnings

On every command the CLI polls GET /api/v1/updates (anonymous, 1h cache in ~/.wonda/state.json) for active announcements: deprecation notices, incident heads-ups, upgrade prompts. Messages are printed to stderr only, so stdout/JSON stays clean for piping.

Per-request deprecation hints arrive as the standard Warning: 299 - "<message>" HTTP header and are surfaced to stderr by the CLI's HTTP client as [deprecated METHOD /path] <message>.

Silence both channels with WONDA_QUIET=1 (env var) or --quiet (flag). Disable just the network checks with WONDA_NO_UPDATE_CHECK=1.

WAB / Wonda Automation Browser (wonda wab)

1Password is included in local WAB profiles. WAB downloads the official stable 1Password browser extension, verifies its publisher signature, and pins it to the toolbar on first setup. Open the WAB with wonda wab show <persona>, then click 1Password to sign in. Each persona keeps its own 1Password session. The desktop app may require you to approve WAB as an additional browser; signing in directly in the extension also works. Cached extension files work offline. WAB checks cached extensions for updates at most once a day when a persona starts, including failed checks; changing the Chromium version triggers another compatibility check. If the extension has never installed successfully, each start retries installation. An already-running browser picks up an extension update on its next start. Startup and wab install prune unused cached versions older than seven days; versions used by a live WAB process are kept. Set WAB_ONEPASSWORD=0 before starting WAB to skip the bundled extension. Anonymous captures, scratch sessions, and cloud twins do not load it. A Wonda data path containing a comma also skips 1Password with a driver-log warning; use a WONDA_HOME path without commas to enable the extension. Enhanced Safe Browsing or administrator policies can also block automatic extension loading; WAB reports Chromium’s rejection in its driver log and preserves your browser settings.

The Wonda Automation Browser (WAB) is a premium stealth antidetect browser, hardened so platforms cannot fingerprint it as automation. wonda wab is the one command for the antidetect Chromium stack (an undetected Playwright fork). It has two faces:

• Authenticated sessions. One persistent headful Chromium per persona that holds signed-in sessions for LinkedIn, X, Reddit, and friends. The CLI spawns it on demand, lets it idle out, and routes platform reads/writes through it whenever a command runs --via wab. Cookies live in the persona's Chromium profile, not in ~/.wonda/config.json.
• Anonymous capture. wonda wab screenshot <url>, wonda wab record <url>, and wonda brand extract drive an ephemeral Chromium with a fresh fingerprint, no persona, and no cookies. Screenshot supports responsive single, batch, and manifest capture. See the screenshot and record blocks below.

The mental model: you have accounts (one identity per platform). Each platform command routes to that account's cookies via either the flat JSON store (--via cookies, fast, no Chromium) or the account's persona (--via wab, live antidetect Chromium). A persona is the Chromium envelope that can hold multiple accounts under one fingerprint. In almost every case the persona is auto-created on first --via wab use, named after the account, so you never type a persona name.

The local wonda.mcpb Desktop Extension uses this same local WAB path from Claude Desktop or Claude Code: platform cookies stay on-device, reads use local cookies, and writes use the local WAB. Claude web and Cowork need the remote MCP connector instead.

Native login is the default for a new persona. wonda wab login <persona> <platform> opens a headful WAB window and you log in there. The session is minted INSIDE the WAB, so it is independent (logging out of the same account in an unrelated Chrome cannot revoke it) and the cookies are born under the WAB's own fingerprint, so session and browser identity stay coherent. A brand-new persona auto-created on first --via wab use chains straight into this flow on a TTY. After an X login, Wonda detects the signed-in screen_name and records it as the persona's X account binding. Existing bindings are never silently changed; a different detected handle produces a warning. Pasting cookies from another browser (wonda linkedin auth set, wonda x auth set, ...) still works and is the explicit fallback, but a hand-pasted li_at on a novel WAB fingerprint is the highest-risk shape.

wonda wab install                             # one-time: npm install + stealth-browser Chromium (shared by sessions, screenshot, record, brand extract)
wonda wab update                              # refresh the runtime to the version this CLI pins (runtime versions ship inside CLI releases): staged download, verified, then atomically swapped in, so a failed download leaves the current driver tree untouched; refuses the runtime swap while personas run (wab stop them first; the relay picks the new runtime up on its next spawn, no restart), while an already-current tree's cheap manifest refresh proceeds without the gate: it reconciles drifted current files back to this build's embedded bytes (additive rewrites, safe under a running persona) and only prunes stale files when NO personas are running, deferring the prune (and keeping the mismatch visible) until the tree is quiet; --check reports installed vs pinned without changing anything, --force reinstalls even when current and re-downloads the shared Chromium build (corrupt-cache recovery; caveat: --force clears the cached Chromium before re-downloading, so a failed forced download leaves the shared cache without it until a retry succeeds). The driver tree also carries a manifest recording what produced it: the embedded driver asset identity compatibility is actually keyed on, plus the CLI build and platform as provenance. A tree whose recorded assets differ from this binary's embedded set (or that predates manifests) is refreshed automatically at the next browser spawn or wab update; two builds shipping byte-identical driver assets share one manifest without churn, so the recorded CLI version can lag behind the newest compatible binary. --check reports the manifest state (ok|absent|mismatch|unreadable) without touching anything; --json only changes the output format and still performs the update
wonda wab start [account]                     # spawn (offscreen by default; --visible to show)
wonda wab stop [account]                      # graceful shutdown
wonda wab show [account]                       # peek a background WAB on-screen to watch it (suspends the macOS focus guard); starts it offscreen first if needed
wonda wab hide [account]                       # send a surfaced WAB back offscreen, resume silent background operation
wonda wab screenshot [persona]                 # persona mode: capture an already-open tab without surfacing it; --json still returns inline base64, --output writes a file, --tab/--full-page optional
wonda wab screenshot <url> --output page.png   # anonymous PNG in a fresh ephemeral browser; responsive waits, injection, animation, element, clip, batch, manifest, and diagnostic JSON controls, see below
wonda wab browse [url] --persona <persona>     # load a page and scroll it like a person: pause, scroll, pause, scroll; stops early once the scrolled element stops advancing, see below (plain text output, not JSON)
wonda wab snapshot -i --persona <persona>      # drive ANY page: interactive elements with refs ([ref=e5]); then click/type/scroll/... @e5, see "Drive any page" below
wonda wab menubar                              # Tray control, macOS menu bar or Windows notification area: the Wonda cat icon with a corner status badge. Green filled = running here and serving; orange half = standing by, which proves only that this machine is not the active device (whether another device or no device is serving is a separate three-way the badge cannot answer, and the menu's detail line only echoes the last relay-health check, a snapshot that can lag while this machine stays in standby); red triangle = running but cannot serve; grey hollow = not running here. The badge carries shape as well as color (filled/half/triangle/hollow) on both platforms; macOS falls back to a 🐱 text item with the same states as glyphs. Click for Restart Relay, a state-following Stop/Start toggle (`wonda relay disable`/`enable`, never `relay stop`), Open Log (macOS only; the Windows task runs the relay with no log file), and Show/Hide per running WAB; the tooltip carries the full status sentence. The bottom item "Quit Wonda" runs `wonda app quit` (stop relay + disable autostart + remove every open-at-login mechanism (macOS login item, Windows Startup entry) + remove the icon; quit stays quit until the app or `wonda app open` runs again). The Windows tray is a PowerShell/WinForms NotifyIcon consuming `relay health --json` (state) plus a periodic `wab status --json` (the persona Show/Hide list) and shelling wonda verbs, zero platform logic in the script. --stop removes only the icon and leaves the relay running
# macOS Dock menu: right-click a running WAB's Dock tile (the 🐱) for "Show on screen" / "Send to background" (same as wab show/hide). Each running persona has its own Dock tile and its menu controls only that persona. Opt out with WAB_DOCK_MENU=0.
# macOS: a background WAB no longer steals focus or flashes the menu bar / Dock when it opens a new tab; the Dock tile stays, it just never comes to the foreground until you `wab show` it.
wonda wab status                              # list personas + last activity + browser-context health (a running daemon whose Chromium died shows browser-dead with a restart hint instead of a misleading plain "running")
wonda wab login <account> <linkedin|x|reddit|instagram> # RECOMMENDED for a new persona: open headful window, user logs in, session minted in-WAB (independent + fingerprint-coherent)
wonda wab check <account> <linkedin|x|reddit|instagram> # non-interactive session-alive probe
wonda wab bind <persona> --x <acct> --reddit <acct> --linkedin <acct>  # multi-account power-user path: bind N accounts to ONE persona
wonda wab record <url>                        # anonymous one-shot webm recording (no account, no cookies), see below
wonda wab sync-cookies [account]              # force wab → disk cookie sync now (don't wait for the 10-min timer)
wonda wab logs [account] --tail 100           # tail driver.log (--audit for structured per-command log)
wonda wab errors --tail 20 --since 24h        # tail the cross-persona action-failure log
wonda wab top-failures --since 7d             # rank local WAB failures by platform/action/reason, joined with DOM recovery stats
wonda wab top-failures --platform x --json    # machine-readable local failure ranking
wonda wab bundle-failures list                # recent action failure bundles (one per failed run: screenshot, dom, visible-elements, cookies-summary REDACTED)
wonda wab bundle-failures show <id>           # print manifest + file tree for a bundle (id = unix-ms-ts prefix)
wonda wab bundle-failures ship <id>           # zip to ~/Downloads/wonda-failure-<id>.zip for sharing
wonda wab bundle-failures prune               # remove bundles older than 30d (or --max-per-persona, --all)
# Telemetry: on every wab action failure we report (action, platform, reason, error-string, has_bundle, cli_version) as a wab_action_failed PostHog event so maintainers can spot platform rotations across users. NO bundle contents, NO cookies, NO DOM, NO screenshots leave the user's machine. Opt out: WONDA_TELEMETRY_DISABLED=1. For server-side breakdowns, group `wab_action_failed` by `platform`, `action`, `reason`, and `has_bundle` in PostHog. Locally, `wonda wab top-failures` reads only `~/.wonda/wab/errors.jsonl` and persona-local `dom-recoveries.jsonl`, then shows count, last seen, recovered rate, bundle count, and a sample bundle id.
wonda wab migrate-legacy                      # copy a legacy WAB-driver profile into a persona slot
wonda wab restore <persona> [timestamp]       # restore from an hourly snapshot (--list to enumerate)
wonda wab backup disable                      # opt out of auto-push (on by default; existing cloud backups untouched)
wonda wab backup enable                       # opt back in (auto-push synced cookie JSON to wondercat after every disk sync)
wonda wab backup status                       # show config + remote inventory
wonda wab backup push [account]               # one-shot manual push for all platform bindings
wonda wab backup pull [account]               # guarded restore to ~/.wonda/<platform>-cookies/<account>.json; refuses non-empty local unless --force
wonda wab backup pull [account] --dry-run     # preview restores without writing
wonda wab backup list                         # inventory of cloud backups, including device/provenance metadata when available
wonda wab backup delete <plat> <persona> [acct] # remove one backup
wonda wab cookies list                        # explicit cookie backup inventory, metadata only
wonda wab cookies status [account]            # local cookie files plus cloud backup rows
wonda wab cookies port <plat> <persona> [acct] --from-device <id|label> # safely port one selected cloud row to this machine
wonda wab config set <persona> <key> <value>  # persist per-persona spawn defaults (idle-timeout, locale, visible, interactive, proxy_url, timezone, geo_lat/lon)
wonda wab config get <persona>                # print a persona's persisted config

Keep cron personas warm. A persona that continuously backs cookie-only read crons can use wonda wab config set <persona> idle-timeout off, followed by wonda wab start <persona>. The WAB then stays up and its existing 10-minute cookie sync keeps the flat files current. Use always-on only for cron-backing personas; wonda wab config unset <persona> idle-timeout restores the default 30-minute idle shutdown.

Local browser proxy (proxy_url). By default the local WAB dials direct (your own IP). Set wonda wab config set <persona> proxy_url managed to route the LOCAL browser through your account's minted twin proxy, so it shares the same egress as the cloud twin (useful for IP continuity or a VPN/office/CGNAT network). A literal socks5://…/https://… value is a manual override instead; unset clears it back to direct. The proxy is optional: if minting is disabled for the environment or unavailable, the browser falls back to a direct dial.

Lifecycle commands take an --account (e.g. wonda wab login <account> linkedin); the persona is auto-derived from the account name. wonda wab bind is the one place a persona is named explicitly: use it when one Chromium must host accounts that have different names per platform.

Drive any page (snapshot, act on @ref, snapshot). Click, type, scroll, and read ANY site in the persona's logged-in WAB, offscreen, without an action script. Use it for every site or step the platform commands do not cover; never drive the WAB with computer-use or screen clicks. The loop:

1. wonda wab goto <url> --persona <p> (or start from whatever the tab already shows).
2. wonda wab snapshot -i --persona <p> prints the actionable elements as YAML, one per line, each with a ref:

   # url: https://news.ycombinator.com/
   # title: Hacker News
   - link "new" [ref=e7]:
     - /url: newest
   - textbox "Search" [ref=e41]
   

3. Act on a ref: wonda wab click @e7, wonda wab type @e41 "rust" --submit.
4. Snapshot again. Refs expire on every snapshot and navigation of the tab; a stale ref fails fast with ref_not_found and a hint to re-snapshot. Add --snapshot to any page-changing action (click, type, fill, press, hover, select, scroll, goto, back, forward, reload, upload) to print the fresh interactive snapshot in the same call.

Target grammar (one argument; quote it when it has spaces): @e5 / e5 / ref=e5 (snapshot ref; @f1e3 inside a frame), label=Email, placeholder=Search, 640,360 (viewport x,y, click and hover only), anything else is a Playwright selector ('text=Sign in', 'role=button[name="Post"]', CSS, xpath=//a). Prefer refs.

wonda wab snapshot [-i] [--max-chars N]                  # YAML accessibility tree; -i = actionable elements only (default cut 40000 chars)
wonda wab click <target> [--button left|right|middle] [--double]   # human mouse path; scrolls off-screen elements into view first
wonda wab type <target> <text> [--clear] [--submit]      # clicks the field, human keystrokes; --submit presses Enter after
wonda wab fill <target> <text>                           # instant value set, replaces content ("" clears)
wonda wab press <key> [--target <target>]                # Enter, Tab, Escape, ArrowDown, PageDown, ControlOrMeta+A (lowercase ok)
wonda wab hover <target>                                 # open hover menus / tooltips
wonda wab select <target> <value> [value...]             # native <select>, by option value or label
wonda wab scroll [up|down|left|right] [pixels]           # mouse wheel, default down 600; prints scrollY/scrollHeight/atBottom
wonda wab scroll --to top|bottom                         # or --into <target> to bring an element into view
wonda wab wait <target> [--state visible|hidden|attached|detached] [--timeout 10s]
wonda wab wait --url <substring|glob> | --load load|domcontentloaded|networkidle | --time 2s   # exactly one condition; --timeout max 60s
wonda wab get text|html|value|attr|url|title|box [target] [--attr NAME] [--max-chars N]   # text/html with no target = whole page; box prints {x,y,width,height}
wonda wab eval <js> | --file <path> | - [--main-world]   # prints the JSON result; function expressions are called, promises awaited
wonda wab goto <url> [--wait-until load|domcontentloaded|networkidle]   # offscreen navigation; http(s) only; bare host gets https://
wonda wab back | forward | reload                         # reload also takes --wait-until
wonda wab upload <target> <file> [file...] [--via-chooser]   # file input without a native picker; --via-chooser clicks a button that opens one
wonda wab tabs [list] | tabs open <tab> [url] | tabs close <tab>
wonda wab screenshot --persona <p> [--target <target>] [--output shot.png]   # look at the page, or crop to one element

Every page command takes --persona (default: configured default account; a persona that does not exist on this machine is an error, never a fresh blank profile) and --tab (default default, the shared tab wonda wab start --open and MCP wab_open show; any other name is created on first use, so --tab research keeps side work off the page the user watches). Output is one short plain-text line per action even when piped (clicked @e5 (url: ...)); snapshot prints YAML, get prints the raw value, eval prints JSON. --json prints the driver response as one JSON line (e.g. {"ok":"click","url":...}), plus a snapshot field with --snapshot. Errors go to stderr with exit 1 and usually a Hint: line naming the next step. Text starting with - goes after --: wonda wab type @e4 -- "-5 degrees".

• type vs fill: type types keystroke by keystroke with human timing, so use it wherever the site watches input (search with autocomplete, post and chat composers, login forms). fill sets the value instantly; use it for long text in plain form fields. Neither submits: add --submit to type, or wonda wab press Enter.
• eval runs in an isolated world by default (sees the DOM, not the page's JS globals, invisible to the site); --main-world reads page globals but is detectable. The result must be JSON. Prefer snapshot/get for reading.
• goto injects the stored LinkedIn/X/Reddit/Instagram session into a cold profile first and warns on stderr (without failing) on session_revoked, rate_limited, or stale_cookies; a captcha or a page that never loads exits 1. Page-changing commands that land on those platforms flush the refreshed cookies to disk like every other WAB action.
• Platform writes (posts, DMs, connects) still go through the platform commands, which carry recipient guards, rate limits, and audit logs. Use page commands for everything around them.

Scroll a page like a person (browse). wonda wab browse [url] --persona <persona> loads a page in the persona's WAB and scrolls it: a pause to look at the page, then scroll, pause, scroll, for --scrolls times (default 5). --scrolls accepts 1-200; a value outside that range is a hard error, not clamped, and it is rejected before a browser is launched. --first-wait (default 10s) is the pause before the first scroll; --wait (default 5s) is the pause between scrolls; both are jittered +/-30% because a precisely repeated interval is itself a fingerprint, and both are floored at 250ms so a very small value is not jittered down to zero — --first-wait 0 --wait 0 still pauses ~250ms per wait, not 0, so with --scrolls 200 that floor alone adds up to roughly 50s. It only scrolls — no clicking, no engagement — so it works on any site.

Progress is measured on the element that was actually scrolled: the viewport-filling overflow container when the site has one (LinkedIn's feed lives in <main id="workspace">, where window.scrollY never moves at all), otherwise the document scroller. Scrolling stops early once that element stops advancing across 2 consecutive scrolls, reported as (reached bottom) in the printed summary, so a short page does not grind against the bottom. An infinite feed keeps going for the full --scrolls.

--persona falls back to your configured default account exactly like the other wonda wab commands, and an invalid persona name is rejected up front instead of silently creating an empty logged-out profile. With a url, browse navigates in its OWN isolated tab (like every other WAB write) and reports the page actually browsed (after any redirect). Omitting the url dispatches to the persona's shared default tab instead — the driver's initial page, the same tab id wonda wab show/hide reference — scrolling whatever that tab already shows. wonda wab login does NOT leave anything there: it opens and navigates its own separate <platform>-login tab, so a persona that was just logged in still has an untouched default tab (often still about:blank). wonda wab show/hide don't navigate the default tab either — they only toggle the window's on/off-screen visibility — so neither one "opens a page" there. What actually leaves a page on default is something that navigates it directly, e.g. wonda wab start --open <platform|url> (also what the MCP wab_open tool calls). This is NOT a way to continue a page from a previous wonda wab browse <url> run: that run navigated in its own isolated tab, which the no-url form never sees. If the default tab has no page open (about:blank) — the common case right after logging a new persona in, since login's tab is separate — the command fails and asks for a url rather than reporting a successful scroll of nothing.

browse prints a plain-text summary, not JSON, so the global --json / --fields / --jq flags do not apply to it.

wonda wab browse https://example.com --persona <persona> --scrolls 6
wonda wab browse --persona <persona>                      # scroll the shared default tab's current page, no navigation

Anonymous PNG capture (screenshot). wonda wab screenshot has a compatibility-preserving persona mode and a new anonymous URL mode.

wonda wab screenshot [persona] still captures the persona's already-open tab without surfacing the window. Its existing flags retain their meaning: --tab selects the tab, --full-page captures the scrollable page, and --output writes a file. With --json and no output file it still returns {path, base64, mimeType} so MCP and existing automation receive the inline PNG unchanged. --persona <name> works in place of the positional (the same flag every page command takes), and --target <target> crops to one element using the page-command target grammar (@e12 from wab snapshot, label=..., or a selector); --target conflicts with --full-page.

An absolute http:// or https:// argument selects anonymous mode. It launches a fresh ephemeral Chromium with no persona, cookies, or persistent state:

wonda wab screenshot \
  'http://127.0.0.1:8765/hero-lab.html#close-colorflight' \
  --output tmp/close-proof.png \
  --viewport 2048x982 \
  --scale 1 \
  --wait-until networkidle \
  --wait-for '#hero-preview' \
  --delay 700ms \
  --animations disabled

Anonymous defaults are --viewport 1280x720, --scale 1, --wait-until networkidle, --delay 0, and --animations allow. --wait-until accepts load, domcontentloaded, or networkidle. Wonda automatically waits for document.fonts.ready; --wait-for <selector> waits for a visible element, and --delay <duration> adds a final settling delay. --inject-js <file> runs after navigation in an async IIFE, so top-level await works.

For deterministic motion, use --animations disabled to finish finite animations and cancel infinite ones at capture, or --freeze-at 800ms to seek Web Animations to an exact timeline offset and pause them. The two options conflict. --selector '.hero-stage' captures the first matching element. --clip x,y,width,height captures an exact document-coordinate rectangle in CSS pixels. The rectangle may extend beyond the viewport but must fit within the rendered page. --selector, --clip, and --full-page are mutually exclusive capture modes.

Pass several absolute URLs to capture all of them. Or pass relative path, query, or quoted hash arguments after one absolute base URL; when any relative target is present, the first URL is only the resolution base and is not captured separately. Add --viewports for a viewport matrix. Chromium launches once for the whole batch:

wonda wab screenshot http://127.0.0.1:8765/hero-lab.html \
  '#fast-close' '#close-colorflight' \
  --viewports 2048x982,1440x1000,390x844 \
  --output tmp/qa \
  --output-template '{route}-{viewport}.png'

For a single capture without --output-template, --output is the PNG path. For a batch, manifest, or any template-driven capture, it is an output directory. --output-template supports {index}, {route}, and {viewport}, and expanded paths must be unique. {route} is sanitized; overlong expanded path components are shortened with a stable hash suffix. --viewport and --viewports conflict. Without output flags, one URL writes screenshot-<timestamp>.png in the current directory and a batch writes under screenshots-<timestamp>/. A template without --output is relative to the current directory, or to the manifest directory when it comes from the manifest.

A strict JSON manifest describes the same matrix:

{
  "url": "http://127.0.0.1:8765/hero-lab.html",
  "routes": ["#fast-close", "#close-colorflight"],
  "viewports": ["2048x982", "1440x1000", "390x844"],
  "scale": 1,
  "waitUntil": "networkidle",
  "waitFor": "#hero-preview",
  "delay": "700ms",
  "injectJs": "scripts/visual-state.mjs",
  "animations": "disabled",
  "outputTemplate": "{route}-{viewport}.png"
}

Run it with wonda wab screenshot --manifest visual-qa.json --output tmp/qa. Use urls instead of url for independent absolute targets; routes resolves relative to url. Manifest keys mirror the anonymous scalar flags in camel case, including freezeAt, selector, clip, and fullPage. Explicit CLI flags override manifest values.

Anonymous --json returns {ok, captures: [...]} without embedding PNG bytes. Every capture reports url, finalUrl, viewport, dpr, pageDimensions, consoleErrors, failedRequests, and fontsLoaded. A successful capture also reports path, which means the PNG was written there. A failed capture omits path, carries error, and keeps any diagnostics collected before the failure. The browser continues the remaining matrix after an item fails and exits nonzero when any capture failed. Use this output for visual QA diagnostics. Non-JSON output prints only successfully generated PNG paths.

Anonymous video capture (record). wonda wab record <url> records a URL to webm in an ephemeral Chromium (fresh fingerprint each call, no persona, no cookies). Use it for cookie-banner-gated pages (Notion public shares, pdf.js renders, any site where bare Playwright trips a bot check) and marketing demo capture.

wonda wab record https://example.notion.site/page \
  --output recording.webm \
  --duration 5 \
  --viewport 960x1080 \
  --inject-js scripts/page-script.mjs   # optional: runs after load, before timer starts

# Transcode webm to mp4 at 30 fps (the stealth browser records webm/VP8)
ffmpeg -y -i recording.webm -t 5 -r 30 -an \
  -c:v libx264 -pix_fmt yuv420p -crf 18 recording.mp4

The --inject-js file is wrapped in an async IIFE so top-level await works. It runs AFTER domcontentloaded + networkidle + 400 ms paint settle, BEFORE the duration timer starts. Any await inside counts against the recording window. Use it for dark-theme injection, cookie-banner removal, scroll animations, anything that needs to happen in page context.

Node.js requirement: wonda needs Node >= v20 on PATH. Brew users get it via the node dependency; npm users have it by definition; install.sh users may need brew install node (or any Node distribution). If Node is missing, wonda wab install fetches a private copy into ~/.wonda/node/.

Cookie cloud backup. On by default (opt out per machine with wonda wab backup disable). The WAB driver pushes the synced cookie JSON for each bound platform to the wondercat backend after every wab → disk sync and graceful shutdown; auto-push no-ops when no api_key is configured. Encrypted at rest server-side (AES-256-GCM) when SOCIAL_COOKIES_KEY is set, else plaintext jsonb; the wire payload is always plaintext because the server holds the key. Cookie values are never printed by list/status/port commands.

Recovery is guarded. wonda wab backup pull <account> and wonda wab cookies port <platform> <persona> [account] refuse to overwrite a non-empty or newer local cookie file unless --force is passed. Forced writes create a hidden .before-pull-* backup first. Use --dry-run to inspect planned writes. When the backend exposes multiple device rows for the same platform/persona/account, use wonda wab cookies port ... --from-device <id|label> so the source row is explicit.

Current backend compatibility: legacy servers still expose one last-write-wins row per (account, platform, persona, account_label). Newer servers may include device_id, device_label, source, status, generation, and provenance; the CLI displays those fields when present and shows legacy rows as device legacy.

Source lives at cli/wondercat/wab/. The driver is launch.mjs and per-platform action scripts under actions/<platform>/.

WAB reads fail early when the selected browser profile has no live platform session cookie. For LinkedIn, X, Reddit, and Instagram, the error includes wonda wab login <persona> <platform> instead of surfacing an unexplained platform 401/403. This preflight is read-only; writes retain their existing error handling and native login itself is unaffected.

Per-command transport (--via). linkedin, x, and reddit commands take:

• --via cookies|wab: cookies reads the flat per-account JSON store (fast, no Chromium); wab routes through the account's persona Chromium (cookies + TLS fingerprint inherit from a real browser session). An unsupported value errors loudly rather than silently downgrading.
• --via public: paid public-data API where a command explicitly supports it. For LinkedIn this avoids logged-in cookies and WAB profile reads, and uses the public scrape task route for wonda linkedin profile and wonda linkedin enrich.
• --account <name>: which on-disk identity to use (cookie filename / persona). Persona resolution is implicit: the first --via wab use auto-creates a persona named after the account and (on a TTY) chains straight into login.

Defaults differ for reads vs writes. Read commands (profile, posts, search, timeline, etc.) default to cookies (direct API), because that path is fast and detection-safe. Write / engagement commands (post, comment, like, follow, connect, message, mute, repost, delete) default to wab, because the cookie-API path triggers anti-abuse heuristics on LinkedIn / X / Reddit at any meaningful volume. Pass --via cookies to a write command if you explicitly want the legacy API path (where the command supports it).

Commands that require --via wab. A few commands have no cookie path and only run through the Wonda Automation Browser: wonda linkedin comment, wonda linkedin reply-comment, wonda linkedin mute, wonda linkedin follow, wonda linkedin edit-post, wonda linkedin edit-comment, wonda linkedin delete-comment, wonda linkedin post --media, wonda x delete, wonda x reply --attach, wonda x dm send, wonda x dm accept, and wonda x dm start. On these, the default already resolves to wab (one stderr line noting it); passing --via cookies explicitly errors. Reddit's writes (vote, comment, subscribe, save, unsave, delete, and subreddit submit) are likewise wab-only.

Where it runs (--engine). --via picks the transport (browser vs. cookies); --engine picks the location, and the two are orthogonal. Values: local (this machine's WAB), cloud (the account's cloud twin, reached through the twin-action API with the control-session warm-up hidden behind a blocking wait), or auto (the default). auto resolves to local when the identity lives on this machine and cloud when it only exists as a cloud twin (a persona with no local footprint, or one cached as home=cloud from wonda twin provision). So wonda linkedin posts <profile> --account <twin> --engine cloud --via wab returns recent posts through the cloud twin's browser with the same structured result as the local command. --via works the same on both engines: reads default to cookies, writes default to wab, and an explicit supported override is preserved. Wired for LinkedIn (posts, connect, like/unlike, comment, reply-comment, edit-comment, send-message, follow, mute, delete-post, edit-post), X (like/unlike, bookmark, retweet/unretweet, follow/unfollow, delete), Reddit (vote, subscribe, save/unsave, delete), and Instagram (comment); other verbs run local only. --engine is accepted at the platform level (so both linkedin --engine cloud posts and linkedin posts --engine cloud work) but honored only on the wired verbs; passing it to another read or an unwired write is a clear error, not a silent no-op. wonda twin run-action is deprecated in favor of <platform> <verb> --engine cloud (it still works so running agents are not broken). On cloud, like supports plain likes and reactions (--reaction, routed to the react action); only comment reactions (--comment) stay local for now. auto resolves to local when a persona has no local footprint and no cloud twin (so first-use auto-create still works), and to cloud when a cloud twin exists for it. An explicit --engine cloud always runs on the cloud twin and never falls back to a live local relay.

Per-account credentials. Cookies live in per-account JSON files on disk:

• ~/.wonda/x-cookies/<account>.json
• ~/.wonda/reddit-cookies/<account>.json
• ~/.wonda/linkedin-cookies/<account>.json (auto-migrated from the legacy single-file format)

Each file is a session-owned local cache, not a portable credential. A platform account may have separate sessions on multiple computers and a cloud Twin. The file records the owning device/persona WAB session, and every injection, overwrite, cookie-only read, refresh, and backup restore checks that identity first. The CLI does not upload these cookies to or download them from the legacy shared hosted-token store. --force never bypasses a session mismatch.

Pass --account <name> to auth set to keep multiple logins side-by-side on the current device. The binding is recorded against the resolved account persona in account-bindings.json, even when --persona is omitted, and if the matching persona's Chromium is running, the rotated cookies get pushed into that live context. Never use auth set to copy cookies from another device or a Twin. Native wab login is safer. The driver also syncs cookies back to disk every 10 minutes (and on graceful shutdown), so rotated cookies (ct0 cycles, token_v2 server-side refresh, etc.) flow back to the cookies path without manual re-paste. Direct X cookie requests also absorb response cookie rotation into the same store and send the full stored cookie jar, preserving device-trust cookies across reads.

The cloud Twin is its own stable, always-on session. Ordinary provisioning does not seed LinkedIn from local cookie files; use wonda twin login <persona> --platform linkedin to log in inside the Twin. Cloud backups are recovery artifacts for their source session only. Inspect local and cloud ownership with wonda wab cookies status [persona]; it never prints cookie values. Recent LinkedIn posts have a direct structured cloud path: wonda linkedin posts <profile> --persona <persona> --engine cloud --via wab. Other reads that are not yet engine-wired can run inside the Twin with wonda twin run-now <persona> --command "<platform read command>"; retrieve the captured result with wonda twin output <twinRunId>, using the id returned by run-now. Neither path downloads Twin cookies locally.

Safely refresh LinkedIn's flat cookie file. Run wonda linkedin auth refresh --account <account> --persona <persona> before a cookie-only read batch. Fresh disk cookies are a fast local no-op. Stale or near-expiry cookies cause one WAB start (a no-op when already running), one WAB-to-disk sync, and a local/WAB-side LinkedIn session check. The refresh path never invokes the raw wonda linkedin auth check probe and never retries credentials. If the WAB session is dead, it exits nonzero with native re-login required; stop the batch and recover manually with wonda wab login <persona> linkedin.

wonda linkedin auth refresh --json returns {fresh, refreshed, ageSeconds, expiresAt, sessionAlive}. fresh describes the final disk state; refreshed is true only when WAB-to-disk sync ran; ageSeconds is the disk cookie age; expiresAt is the known expiry or null; and sessionAlive is the local/WAB-side login result. sessionAlive is null on a fresh local no-op because no WAB check was needed. --fresh-within changes both the maximum accepted disk age and the minimum remaining recorded li_at lifetime (15 minutes by default). Treat a nonzero exit as authoritative even with JSON output.

Cookie-backed LinkedIn reads accept opt-in --freshen alongside --via cookies. It runs the same local preflight only when the disk cookies are stale, then attempts the requested read once. The default remains unchanged: without --freshen, --via cookies stays fast and browser-free.

Safely refresh X's flat cookie file. wonda x auth refresh --account <account> --persona <persona> uses the same local-first lifecycle with X-specific freshness rules. Since auth_token has no locally readable expiry claim, freshness is the cookie store age only. A store newer than --fresh-within (15 minutes by default) is a browser-free no-op. A stale store starts the selected WAB if needed, syncs X cookies to disk, and checks the WAB-side session without invoking the raw x auth check probe. A dead session exits nonzero with the exact native login command.

wonda x auth refresh --json returns {fresh, refreshed, ageSeconds, sessionAlive}. Cookie-backed X reads accept opt-in --freshen; it runs the same refresh only when the selected disk store is stale, pins the read to the refreshed account, and never applies to WAB reads, writes, or auth commands.

Use wonda x auth status --account <name> for a pure-local view of cookie origin, latest capture provenance, generation, ownership, cookie names, and missing device-trust-cookie risk. It never contacts X and never prints cookie values. Use wonda x browser-bootstrap --account <name> to explicitly push the selected stored jar into running WAB personas bound to that account.

Action rate limits

Every platform command (linkedin, x, reddit, instagram), reads AND writes, runs through a per-profile rate-limit guard so a burst doesn't trip a platform's shadow-ban / anti-abuse heuristics. Accounting is per (platform, account) in a rolling 24h window, logged per profile under ~/.wonda/wab/personas/<persona>/ (so a cloud twin's caps persist across runs).

• Reads are paced, never blocked: spacing is jittered to keep a profile under read_per_min (75/min default), holding across separate invocations.

• Writes are checked against per-bucket daily caps. The LinkedIn defaults (other platforms track + count toward the total/day but have no per-type write cap by default):

  bucket	commands	safe	max
  outreach	connect + send-message + inmail	20	40
  post	post	3	5
  comment	comment, reply-comment	10	20
  react	like	25	50
  visit	visit	15	30
  search	search, search-posts	25	50
  total (all non-read)	any of the above	warn at 90%	100

salesnav search (Sales Navigator) is exempt: it carries no Commercial Use Limit, so it paces as an uncapped read instead of counting toward the search cap or the total/day.

Caps are SOFT by default: an over-safe / over-max action prints a shadow-ban-risk warning to stderr and proceeds. Pass --hard (or set mode: hard in config) to make over-cap writes abort (exit 1) instead.

wonda actions is a JSON data query (not a dashboard) for reading a profile's rolling-24h usage vs caps on demand; the caps/pacing/warnings run silently in the live hook regardless.

wonda actions                        # rolling-24h usage per profile vs caps, as JSON
wonda actions --persona <persona>        # one profile
wonda actions --platform linkedin    # filter to one platform
wonda linkedin quota                 # planning view: remaining LinkedIn allowance per bucket (used vs safe/max + remaining, incl. untouched buckets) BEFORE a big pull, instead of hitting the mid-output warnings
wonda actions sync                   # flush local action/health events to your Wonda account
wonda actions sync --persona <persona>   # flush one profile's ledgers
wonda linkedin post "…" --hard       # enforce caps as hard limits for this command

When an API key is configured, the local ledgers (actions log, WAB audit/error logs, cookie provenance) also sync to your Wonda account-health record automatically in the background on every command: best-effort, batched, and idempotent (a stable client event id per record means retries never double count), so offline use keeps working and sync catches up later. wonda actions sync forces a full flush and prints the server's insert/dedup counts; without an API key it is a silent no-op. Only event metadata travels, never cookie values or failure bundles. WONDA_TELEMETRY_DISABLED=1 turns the background sync off.

Override / disable / hard-mode via ~/.wonda/config.json under action_limits (caps are clamped to safety floors/ceilings so an override can loosen but not silently disable the guard):

{
  "action_limits": {
    "mode": "hard",
    "read_per_min": 75,
    "total_per_day": 100,
    "buckets": { "linkedin": { "outreach": { "safe": 15, "max": 30 } } }
  }
}

Config keys

wonda config get|set|list keys:

• api-key: your wondercat API key.
• base-url: API base (defaults to prod, set to https://staging.api.wondercat.ai for staging).
• default-account: account used when a platform command doesn't pass --account.
• wab-backup-enabled: true/false for cookie cloud backup (same as wonda wab backup enable/disable). On by default; only an explicit false disables it.

Transport is NOT a config key. Each command picks it per kind (reads default to cookies, writes / engagement default to wab), identically on every platform. Override it per command with --via cookies|wab (where the platform supports it).

How to think about content creation

You are a marketing director with access to a full production toolkit. Before touching any tool, think:

1. What product category? (beauty, food, tech, fashion, fitness, etc.)
2. What format performs for this category? (UGC memes for everyday products, cinematic for luxury, before/after for transformations, testimonial for services)
3. What's the hook? (relatable scenario, surprising twist, aspirational lifestyle, social proof)
4. What specific scene? (not "product on table" but "person discovering the product in a funny situation")

Decision flow

When asked to create content, follow this order:

Step 1: Gather context

wonda brand                                                    # Active brand: identity, colors, fonts, logos, products
wonda brand list                                               # All brands owned by this account/org
wonda brand show <brand-id>                                    # Specific brand
wonda brand extract https://stripe.com                         # Local-only: writes ./output/stripe.com/{DESIGN.md, tokens.json, assets/}
wonda brand extract https://stripe.com --save --make-active    # Local + persist + activate (the common path)
wonda brand extract https://stripe.com --save --name "Stripe"  # Persist with a custom name
wonda brand extract https://stripe.com --no-output --save      # Don't write to disk, persist only
wonda brand save                                               # Persist the most recent ./output/<domain>/ dir to the server
wonda brand save --from ./output/stripe.com --make-active
wonda brand pull <brand-id>                                    # Download a saved brand back to ./output/<domain>/
wonda brand activate <brand-id>                                # Set as the active brand
wonda brand upload-logo <brand-id> https://acme.com/logo.svg   # Attach a logo by URL (--variant wordmark|icon|dark|light)
wonda brand upload-font <brand-id> https://acme.com/Geist.woff2 --weight 700
wonda brand delete <brand-id>
wonda analytics instagram                                      # What content performs well
wonda scrape social --handle @competitor --platform instagram --wait  # Competitive research (if relevant)

# Cross-platform research (if relevant)
wonda x search "topic OR keyword"                              # Find conversations on X/Twitter
wonda x user-tweets @competitor                                # Competitor's recent tweets
wonda reddit search "topic" --sort top --time week             # Reddit discussions
wonda reddit feed marketing --sort hot                         # Subreddit trends
wonda linkedin search "topic" --type COMPANIES                 # LinkedIn company/people research
wonda linkedin profile competitor-vanity-name                  # LinkedIn profile intel

Step 2: Check content skills

Content skills are step-by-step guides for common content types. Each skill tells you exactly which models, prompts, and editing operations to use — and in what order. ALWAYS check skills before building from scratch.

Skills are server-hosted, per-account, and editable (the same model as wonda brand save) — you don't download a folder of .md files you own. Wonda ships a canonical set of default skills, served read-only as the fallback. Pull them live for a task; fork a default into your own copy when you want to tweak it; Wonda keeps full version history. skill list shows your effective skills (defaults overlaid with your own edits), and flags any fork whose default has since changed.

wonda skill list                                # Browse your effective skills (defaults + your own); forks with a changed default are flagged
wonda skill get <slug>                          # Pull a skill's full step-by-step guide live to stdout

Default skill catalog (live source: wonda skill list, which also shows your own forks/edits and flags drift):

video

image

social-research

strategy

utility

Editing skills (optional). When a default doesn't quite fit, fork and edit it instead of working around it. Editing a default forks it into your account automatically:

wonda skill create my-ugc --from ugc-talking    # Fork a default into your own editable copy
wonda skill edit my-ugc --editor                # Record a new version (opens $EDITOR; or --file <md> / stdin)
wonda skill diff <slug>                          # See what changed in the default since you forked it (drift)
wonda skill refactor <slug> --editor             # Re-base your fork onto the updated default, clears the drift hint

If a skill matches → wonda skill get <slug>, read it, adapt to context, execute each step.

If no skill matches → build from scratch (Step 3).

Step 2.5: Decide whether finishing should be local

Not every media task should go back through Wonda editing. Use this routing rule:

• Use wonda for AI generation, AI transcription/alignment, scraping, publishing, hosted transitions, and workflows that need media IDs or remote jobs.
• Use local ffmpeg for deterministic transforms on files you already have or can download: trim, crop/scale/pad, concat (merging multiple clips), replace audio, extract audio/frame, reverse, normalize for delivery, burn captions, split scenes, cut silence, and build analysis artifacts. Always merge clips locally — server-side merge can hang for 30+ minutes once any input exceeds ~7MB.

When a task starts from a Wonda media ID but the actual edit is deterministic, move it to local files first:

wonda media download <mediaId> -o ./input.mp4

Before any local ffmpeg work:

which ffmpeg
which ffprobe
ffmpeg -version
ffprobe -v error -show_format -show_streams -of json ./input.mp4

Font rule for local caption/text work:

• Prefer an explicit font file path over a family name.
• Never assume a font exists. Check first with fc-match, fc-list, /System/Library/Fonts, /Library/Fonts, ~/Library/Fonts, or /usr/share/fonts.
• If the task is mainly local finishing/captions/formatting/splitting/artifact extraction, check the ffmpeg skill before inventing commands.
• wonda edit video runs a local ffmpeg for every editor op: trim, crop, volume, speed, reverseVideo, extractFrame, extractAudio, editAudio, imageCrop, imageToVideo, merge, overlay, splitScreen, splitScenes, skipSilence. The render runs on your machine via ffmpeg: no server-side editor_job and no credit hold for the render itself (inputs are downloaded and the result uploaded around it). textOverlay and animatedCaptions also run locally, via the bundled hyperframes (Chromium) renderer. ffmpeg must be on PATH (wonda doctor verifies). The public API /video/edit, /image/edit, /audio/edit are no longer used for these and return 410 Gone.
• Always merge clips locally. Server-side merge can hang for 30+ minutes once any input exceeds ~7MB, and wonda edit video --operation merge now runs in local ffmpeg by default for the same reason.
• Never mix per-clip audio then concat. Concat the video tracks first, then layer the full voiceover or music track once over the joined timeline. Per-clip audio bakes create cut-line collisions and silent gaps.

Default local export target unless the user asked otherwise:

-c:v libx264 -preset medium -crf 18 -pix_fmt yuv420p -movflags +faststart -c:a aac -b:a 192k

Always pass -y as the first flag so the command auto-overwrites the output. ffmpeg prompts interactively when the output path exists and agent shells hang on that prompt until timeout.

Step 2.6: Pick the right local tool

Editing maps to one of four tools. Pick the first row that matches.

Run wonda doctor once on a new machine to confirm ffmpeg, node, and hyperframes are all available. Pass --warm-chrome to pre-fetch hyperframes' bundled Chromium (~150 MB) so the first clipping render doesn't pause to download it. Pass --wab to also audit the on-disk WAB browser runtime (installed driver vs this build's pin, driver tree manifest); offline, read-only, and advisory like --relay.

Examples:

Primitive trim and merge (wonda edit, local ffmpeg):

wonda edit video --operation trim --media $VID \
  --params '{"trimStartMs":3000,"trimEndMs":10000}' \
  --wait -o ./trimmed.mp4

wonda edit video --operation merge --media $A,$B,$C \
  --wait -o ./merged.mp4

Motion graphics intro (wonda compose, hyperframes):

wonda compose motion --template fade-in \
  --text "Q4 Recap" --subtitle "Wondercat" \
  --duration 4 --resolution portrait -o intro.mp4

wonda compose text --input ./clip.mp4 --text "NEW DROP" \
  --position bottom-center -o overlay.mp4

Kinetic captions on a finished clip (transitions service):

wonda transitions run --media $VID --preset caption_word_pop --wait -o final.mp4

Raw ffmpeg for an op no primitive covers (e.g. concat with audio fade out):

ffmpeg -y -f concat -safe 0 -i list.txt \
  -af "afade=out:st=29:d=1" \
  -c:v libx264 -crf 18 -pix_fmt yuv420p \
  -c:a aac -b:a 192k out.mp4

Multi-step pipeline (compose intro → wonda merge with main → transitions captions):

wonda compose motion --template scale-pop --text "Hello" --duration 3 -o intro.mp4
wonda edit video --operation merge --media $(wonda media upload intro.mp4 --quiet),$MAIN_VID \
  --wait -o merged.mp4
MERGED_ID=$(wonda media upload merged.mp4 --quiet)
wonda transitions run --media $MERGED_ID --preset caption_word_pop --wait -o final.mp4

Step 3: Build from scratch (chain endpoints)

When no skill matches, chain individual CLI commands. Each step produces an output that feeds into the next.

Single asset:

wonda generate image --model gpt-image-2 --prompt "..." --aspect-ratio 9:16 --wait -o out.png
# --params '{"quality":"high"}' — auto/low/medium/high (default auto)
# --negative-prompt "..."       — override what to exclude (model-dependent)
# --seed <number>               — pin the seed for reproducible results (model-dependent)
wonda generate video --model seedance-2 --prompt "..." --duration 5 --params '{"quality":"high"}' --wait -o out.mp4
wonda generate text --model <model> --prompt "..." --wait
wonda generate music --model suno-music --prompt "upbeat lo-fi" --wait -o music.mp3

Audio (speech, transcription, dialogue):

# List available voices (TTS + dialogue use the same set)
wonda audio voices

# Text-to-speech
wonda audio speech --model elevenlabs-tts --prompt "Your script here" \
  --params '{"voiceId":"hpp4J3VqNfWAUOO0d1Us"}' --wait -o speech.mp3
# elevenlabs-tts always requires a voiceId — pick one from `wonda audio voices`

# Transcribe audio/video to text
wonda audio transcribe --model elevenlabs-stt --attach $MEDIA --wait

# Multi-speaker dialogue (each speaker needs a voiceId from `wonda audio voices`)
wonda audio dialogue --model elevenlabs-dialogue \
  --prompt 'ALICE: Hi! BOB: Hello!' \
  --params '{"speakers":[{"label":"ALICE","voiceId":"hpp4J3VqNfWAUOO0d1Us"},{"label":"BOB","voiceId":"IKne3meq5aSn9XLyUdCD"}]}' \
  --wait -o dialogue.mp3

Audio AI operations (direct-inference, NOT editor ops):

# Denoise / dereverberate speech
wonda audio enhance --model replicate-resemble-enhance --attach $MEDIA \
  --params '{"denoise":true,"chunkSeconds":10}' --wait -o enhanced.wav

# Split a track into voice and instrumental stems
wonda audio extract-voice --model replicate-demucs --attach $MEDIA \
  --wait -o vocals.wav

Add animated captions to a video:

The animatedCaptions operation handles everything in one step — it extracts audio, transcribes for word-level timing, and renders animated word-by-word captions onto the video.

# Generate a video with speech audio
VID_JOB=$(wonda generate video --model seedance-2 --prompt "..." --duration 5 --aspect-ratio 9:16 --params '{"quality":"high"}' --wait --quiet)
VID_MEDIA=$(wonda jobs get inference $VID_JOB --jq '.outputs[0].media.mediaId')

# Add animated captions (single step)
wonda edit video --operation animatedCaptions --media $VID_MEDIA \
  --params '{"fontFamily":"TikTok Sans SemiCondensed","position":"bottom-center","sizePercent":80,"strokeWidth":2.5,"fontSizeScale":0.8,"highlightColor":"rgb(252, 61, 61)"}' \
  --wait -o final.mp4

The video's original audio is preserved. Do NOT replace the audio with TTS — Sora already generated the speech.

Transitions (effects pipelines on a single video):

wonda transitions presets                            # List built-in presets (JSON)
wonda transitions operations                         # Grouped by category (analysis/effect/...)
wonda transitions operations --json                  # Full per-param metadata
wonda transitions llms                               # Full reference (presets + ops + dependencies)
wonda transitions run --media $VID --preset flash_glow --wait -o out.mp4
# Or send an agent-generated timeline of clips (inline JSON):
wonda transitions run --media $VID \
  --clips '[{"layer_type":"video","start_frame":0,"end_frame":60}]' --wait -o out.mp4
# Or from a file (handy for long agent timelines):
wonda transitions run --media $VID --clips ./timeline.json --wait -o out.mp4
# To attach scene_transitions: pass an envelope (clips + scene_transitions)
# instead of a bare clip array — same file, both fields forwarded.
wonda transitions run --media $VID --clips ./timeline_with_transitions.json --wait -o out.mp4
# where timeline_with_transitions.json is:
#   { "clips": [...],
#     "scene_transitions": [{"name":"crossfade","params":{"duration":8},"boundaries":[60]}] }
wonda transitions job <jobId>                        # Poll a transition job

Use exactly one of --preset or --clips. Requires a full (logged-in) account. Always read wonda transitions llms first when composing a clips timeline. It documents the detect/segment/effect dependencies, which ops need masks, and the full clip-spec shape (layer types, tracks, effects, transforms).

Preset variables (variables block). Each preset declares the template variables it accepts under variables in wonda transitions presets. Each entry has name, description, and required. Required variables MUST be supplied or the job is rejected with a 400 — no more silent skipping. Pass them with --var name=value (repeatable) or, for the common prompt case, the --prompt shortcut:

# flash_glow_prompted requires { prompt }
wonda transitions run --media $VID --preset flash_glow_prompted \
  --prompt "woman in white dress" --wait -o out.mp4

# text_behind_person requires { prompt, text }
wonda transitions run --media $VID --preset text_behind_person \
  --var prompt="the person" --var text="HELLO WORLD" --wait -o out.mp4

# Numeric-typed vars: bare digits are decoded as numbers, "true"/"false" as
# bools, everything else stays a string. Presets that compare frame indices
# numerically (border_frame, marquee_text, quick_motion_text, bg_remove_scale)
# need this — quoting an int turns it back into a string.
wonda transitions run --media $VID --preset border_frame \
  --var exit_start_frame=200 --var exit_end_frame=251 --wait -o out.mp4

The prompt variable is a detection text query describing which subject to mask, fed to SAM3 to produce per-frame segmentation masks. Not a content-generation prompt.

Building a custom --clips timeline that needs detection masks? Add a clip with layer_type: "video" and a mask: {layer_type: "mask", analysis_steps: [{name: segment, params: {prompt: "..."}}]}. SAM3 handles both detection and segmentation in one step from the prompt, so no separate detect step is needed.

Pre-warming masks before render (recommended)

For presets with mask:<label> variables, run wonda transitions ensure-masks first so the render starts with masks already prepared. The first call for a (media, label) pair takes 1-3 minutes; subsequent calls are near-instant.

# 1. Ensure masks are prepared for the labels you'll use, blocking until ready.
wonda transitions ensure-masks --media $VID --labels person,phone --wait

# 2. Run the render. Masks are already prepared.
wonda transitions run --media $VID --preset slide_reflect_background \
  --var "masks=mask:person+phone" --wait -o out.mp4

ensure-masks flags:

• --media MEDIA_ID — required, the video the masks are for
• --label NAME — repeatable, one label per call (--label person --label phone)
• --labels NAME,NAME — comma-separated alternative (--labels person,phone)
• --wait — block until every label is prepared
• --timeout DUR — cap wait time when --wait is set (default 10m)

Multi-prompt syntax: mask:woman+phone in --var is split into separate masks (woman, phone) and unioned per-frame. Pass each sub-label separately to ensure-masks so all of them are pre-warmed.

When to skip ensure-masks:

• Non-mask presets (no mask:<label> variables) — nothing to prepare
• A previous render already used these (media, labels) — already prepared

When ensure-masks matters most:

• First render of a new media with mask-based presets
• Iterating params on a render — pre-warm once, then run as many times as you want without re-preparing

Multi-scene presets (requiresMultiScene: true). Some presets use scene-aware logic and expect a video with multiple cuts/scenes. Check requiresMultiScene in wonda transitions presets. If true, feeding a single continuous shot will produce only one scene and the effect may look underwhelming. Combine clips first or use a video with natural cuts.

Tweaking preset params. Every preset is clip-shape. Pull a single preset with wonda transitions preset <name> --json, read its clips: (single-track) or tracks: (multi-track) field, edit any clip param, and submit as --clips. For multi-track presets, flatten by giving each clip a track index drawn from the track it came from. If the preset declares sceneTransitions:, pass that array through unchanged on the request.

# Single-track preset (e.g. flash_glow_montage): copy clips: directly
wonda transitions preset flash_glow_montage --json | jq '.preset.clips' > clips.json
# edit clips.json
wonda transitions run --media $VID --clips "$(cat clips.json)" --wait -o out.mp4

Auto-repair safety net (--auto-repair, --face-bbox). For --clips renders the worker runs a deterministic repair pass on the submitted JSON before rendering, default on. Repairs: width-fit font clamp, descender clamp against canvas bottom, stack-spacing snap (ROW1_py from cap-height formula), keyframe-bound clamp to [0, source_duration], same-y-row caption overlap trim, mask full-duration extension, stroke-width zeroing, letter-spacing target snap per font, mask-cutout duration extension, negative-start clamp, and (with --face-bbox) face-overlap caption shift. Pass --auto-repair=false for strict validation; out-of-spec values then surface as render errors.

# Push body captions off the speaker's face. bbox is x1,y1,x2,y2 in canvas pixels (top-left origin).
wonda transitions run --media $VID --clips ./timeline.json \
  --face-bbox 200,160,520,520 --wait -o out.mp4

# Strict mode — disable auto-repair to see exactly which clips fail validation.
wonda transitions run --media $VID --clips ./timeline.json \
  --auto-repair=false --wait -o out.mp4

--face-bbox only shifts body captions. Decorative text you want behind the speaker still routes through an explicit mask_cutout {prompt: "person"} clip.

Output URL paths differ by job type:

• Inference jobs (generate, audio): .outputs[0].media.url and .outputs[0].media.mediaId
• Editor jobs (edit): .outputs[0].url and .outputs[0].mediaId

Model waterfall

Image

Default: gpt-image-2. OpenAI's flagship — strongest prompt adherence, best text-in-image, high-fidelity edits via reference images. Handles 1-4 reference images. Quality tiers: auto (default), low, medium, high — pass via --params '{"quality":"high"}'. Caps at 1536px output.

For img2img editing specifically (change, add/remove, restyle, bg-remove, crop, text overlay, vectorize), use wonda skill get image-edit — it has the full edit-specific decision tree.

Pick something else only when one of these applies:

• User explicitly requests another model
• More than 4 reference images → nano-banana-2 (gpt-image-2 caps at 4 refs; nano-banana-2 accepts up to 14). For 1-4 refs, stay on gpt-image-2.
• Need vector output → runware-vectorize
• Need background removal → birefnet-bg-removal
• Cheapest possible / fastest drafts → z-image
• Need >1536px / true 4K output → nano-banana-pro (1K/2K/4K) or nano-banana-2 (1K/2K/4K). gpt-image-2 caps at 1536px.
• gpt-image-2 unavailable / OpenAI down → nano-banana-2 or seedream-4-5 or grok-imagine-pro

Video

Default: seedance-2 (duration 5/10/15s, default 5s, quality: high). Escalation:

• Quality complaint or different style → sora2 or sora2pro
• Max single-clip duration is 15s for Seedance 2, 20s for Sora → for longer content, stitch multiple clips via merge
• Veo (veo3_1, veo3_1-fast) is available but NOT in the default waterfall. Only pick Veo when the user explicitly asks for Veo by name.
• Gemini Omni (gemini-omni-video) is available but NOT in the default waterfall. Only pick it when the user asks for Gemini by name, or specifically needs multi-image reference T2V/I2V (up to 7 reference images) or 4K output.

Image-to-video routing (MANDATORY when attaching a reference image):

• Person/face visible in the reference image → MUST use kling_3_pro (preserves identity better for faces)
• No person in reference image → use seedance-2
• Text-to-video (no reference image): Seedance 2 generates people fine. This rule ONLY applies when you --attach an image.

Kling model family:

• kling_3_pro — Text-to-video and image-to-video, supports start/end images, custom elements (@Element1, @Element2), 3-15s duration, 16:9/9:16/1:1
• kling_2_6_pro — General purpose, 5-10s, 16:9/9:16/1:1, text-to-video and image-to-video
• kling_2_6_motion_control — Motion transfer: requires both a reference image AND a reference video, recreates the video's motion with the image's appearance
• kling2_5-pro — Budget Kling option, 5-10s, supports first/last frame images

Kling prompt rules (important): Kling's prompt field caps at 2,500 characters and Kling responds poorly to Sora-style structured briefs (SCENE: / SUBJECT: / MOTION: / BANNED LOOK: section headers). In that format Kling latches onto atmosphere nouns and silently drops the central subject (verified empirically: the same 2,842-char Sora-style prompt that rendered correctly on Sora 2 Pro and Seedance 2 produced no phone at all on Kling — even when trimmed to 2,250 chars). When escalating Seedance → Kling, or targeting Kling directly, rewrite the prompt as short natural-language prose (~1,000–1,500 chars) and lead with the hero subject in the opening sentence rather than burying it inside a SUBJECT: block. Do NOT pass a Sora-formatted prompt through to Kling unchanged.

Other video models:

• grok-imagine-video — xAI video generation, 5-15s, supports 7 aspect ratios including 4:3 and 3:2
• gemini-omni-video: Google Gemini Omni. Text-to-video and image-to-video with up to 7 reference images (slots reference_image_1 through reference_image_7). Durations 4/6/8/10s, aspect ratios 9:16 and 16:9, resolutions 720p / 1080p / 4K. Pricing: $0.15 base + $0.075/s at 720p/1080p, $0.75 base + $0.075/s at 4K. No native audio (pair with a separate audio model if speech is needed).
• topaz-video-upscale — Upscale video resolution (1-4x factor, supports fps conversion)
• sync-lipsync-v2-pro — Legacy lipsync for user-supplied video + audio pairs. Inferior to native-audio generation and almost never the right choice for new content. See the "Lip sync" section for rules.

Seedance family (DEFAULT video model, watermarks automatically removed):

• seedance-2 — Base Seedance 2.0 (T2V/I2V, 5-15s, high=standard/basic=fast)
• seedance-2-omni — Multi-reference generation (images, audio refs)
• seedance-2-video-edit — Edit existing video via text prompt

Video durations: Accepted --duration values vary by model. Check with wonda capabilities or wonda models info <slug>.

Audio

• Music: suno-music (set --params '{"instrumental":true}' for no vocals)
• Text-to-speech: elevenlabs-tts — only for explicit narrator/voice-over asks over silent footage. Do NOT use to "make a UGC character talk" — Sora / Sora 2 Pro / Veo 3.1 / Kling 3 / Seedance 2 generate native synced speech in any language, which looks and sounds far better. Always set voiceId in params. Default female voice: --params '{"voiceId":"21m00Tcm4TlvDq8ikWAM"}' (Rachel).
• Transcription: elevenlabs-stt
• Multi-speaker dialogue: elevenlabs-dialogue
• Enhance audio (clean up noisy speech): replicate-resemble-enhance via wonda audio enhance — denoise + dereverberate. Use when a voice recording sounds muffled, echoey, or has background noise. NOT a general "sounds better" button; if the source is already clean this can soften it.
• Extract voice (isolate vocals / split stems): replicate-demucs via wonda audio extract-voice — splits into voice and instrumental tracks. Use to pull a speaker or singer off a track, or to isolate the music behind a vocal.

Native synced speech (preferred over TTS + lipsync): Sora, Sora 2 Pro, Veo 3.1, Kling 3, and Seedance 2 all generate dialogue in any language directly inside the video, with mouth movements baked in. Put the line (and language) in the video model's --prompt. Never chain elevenlabs-tts → sync-lipsync-v2-pro to fake speech over a silent generation.

Characters

Characters are reusable saved combos (image + optional voice audio) you can mention in prompts with @name. The server auto-injects the image, optional face video, and audio into the right slots for the selected model. Works on Kling 3 Pro (start_image + element_1 + voice_audio) and Seedance 2 Omni (ref_image_1 + ref_video_1 + ref_audio_1). Name rules: must start with a letter, 1–31 chars, alphanumeric + _/-.

Provider gotchas (Seedance 2 Omni): when a character is mentioned, the API routes Seedance to MuAPI automatically. Replicate enforces a 15s ref_audio_1 cap and rejects famous-celebrity refs with E005 — input flagged as sensitive. MuAPI is the reliable path for character-driven jobs. Even on MuAPI, top-tier celebrity refs (think Sydney Sweeney, Leonardo DiCaprio) are blocked with "Face detected in uploaded image. Please use an image without real people." Non-celebrity faces and lesser-known public figures pass cleanly. If you see that error on a real-person ref, use Kling 3 Pro instead (its character pipeline runs voice cloning server-side, so the raw face audio never touches a moderation classifier).

From a Kling clip — extract a frame + voice from a generation you like:

VID=$(wonda generate video --model kling_3_pro --prompt "young man, grey tshirt, talking to camera" --wait --quiet)
VID_MEDIA=$(wonda jobs get inference $VID --jq '.outputs[0].media.mediaId')
wonda character from-media alex --source $VID_MEDIA --frame-ms 2500
wonda generate video --model kling_3_pro --prompt "@alex welcomes viewers to the channel" --wait -o alex-welcome.mp4

From scratch — generate a portrait and a TTS sample, then bind them:

IMG=$(wonda generate image --model nano-banana-2 --prompt "young woman, studio portrait" --wait --quiet)
IMG_MEDIA=$(wonda jobs get inference $IMG --jq '.outputs[0].media.mediaId')
AUD=$(wonda audio speech --model elevenlabs-tts --prompt "Hi, this is me" --params '{"voiceId":"21m00Tcm4TlvDq8ikWAM"}' --wait --quiet)
AUD_MEDIA=$(wonda jobs get inference $AUD --jq '.outputs[0].media.mediaId')
wonda character create maya --image $IMG_MEDIA --audio $AUD_MEDIA

List / inspect / update / delete: wonda character list, wonda character get <name>, wonda character update <name> --audio $NEW, wonda character delete <name>. Only one character with audio can be referenced per generation.

Prompt writing rules

Follow this waterfall top-to-bottom. Use the FIRST matching rule and stop.

1. PASSTHROUGH — If the user says "use my exact prompt" / "verbatim" / "no enhancements" → copy their words exactly. Zero modifications.

2. IMAGE-TO-VIDEO — When a source image feeds into a video model, describe MOTION ONLY. The model can see the image. Do NOT describe the image content.

   • Good: "gentle breathing motion, camera slowly pushes in, atmospheric lighting shifts"
   • Bad: "Two cats on a lavender background breathing softly" (describes the image)

3. EMPTY PROMPT (from scratch) — Use the user's exact request as the prompt. Do NOT add style descriptors, lighting, composition, or mood.

   • User says "create an image of a cat with sunglasses" → prompt: "create an image of a cat with sunglasses"
   • Do NOT enhance to "A playful orange tabby wearing oversized reflective sunglasses, studio lighting, shallow depth of field"

4. NON-EMPTY PROMPT (adapting a template) — Keep the structure and style, only swap content to match the user's request. Keep prompts literal and constraint-heavy.

Aspect ratio rules

Three cases, no exceptions:

1. User specifies a ratio → use it: --aspect-ratio 16:9
2. User doesn't mention ratio → explicitly set --aspect-ratio 9:16 for social content (UGC, TikTok, Reels, Stories). Portrait is the default for any social/marketing video.
3. Editing existing media → use --aspect-ratio auto to preserve source dimensions

UGC and social content is ALWAYS portrait (9:16). If someone asks for a TikTok, Reel, Story, or UGC video, always use --aspect-ratio 9:16. Landscape is only for YouTube, presentations, or when explicitly requested.

Square (1:1) is supported by all Kling models and some image models — use for Instagram feed posts when requested.

Common chaining patterns

These patterns show how to compose multi-step pipelines by chaining CLI commands. Each step's output feeds into the next.

No need to download and re-upload between steps. Every generation and edit produces a media ID in its output. Pass that ID directly to the next command via --media or --audio-media. Use --jq '.outputs[0].media.mediaId' for inference jobs and --jq '.outputs[0].mediaId' for editor jobs. Only use -o <file> on the FINAL step to download the finished output.

Animate an image to video

MEDIA=$(wonda media upload ./product.jpg --quiet)
# No person in image → Seedance 2
wonda generate video --model seedance-2 --prompt "camera slowly pushes in, product rotates" \
  --attach $MEDIA --duration 5 --params '{"quality":"high"}' --wait -o animated.mp4
# Person in image → Kling (ONLY when attaching a reference image with a person)
wonda generate video --model kling_3_pro --prompt "the person turns and smiles" \
  --attach $MEDIA --duration 5 --wait -o person.mp4

Replace audio on a video (TTS voiceover or music)

# Generate TTS
TTS_JOB=$(wonda audio speech --model elevenlabs-tts --prompt "The script" \
  --params '{"voiceId":"21m00Tcm4TlvDq8ikWAM"}' --wait --quiet)
TTS_MEDIA=$(wonda jobs get inference $TTS_JOB --jq '.outputs[0].media.mediaId')
# Mix onto video (mute original, full voiceover)
wonda edit video --operation editAudio --media $VID_MEDIA --audio-media $TTS_MEDIA \
  --params '{"videoVolume":0,"audioVolume":100}' --wait -o with-voice.mp4

Only use this when you need to REPLACE the video's audio. Sora, Sora 2 Pro, Veo 3.1, Kling 3, and Seedance 2 all generate native synced speech in any language — don't replace it with TTS unless the user explicitly asks for a different voiceover. Never reach for this step to "add speech" to a UGC/talking-head clip; put the dialogue in the video model's prompt instead.

Add static text overlay

Static overlays (meme text, "chat did i cook", etc.) use smaller font sizes than captions. They're ambient, not meant to dominate the frame.

wonda edit video --operation textOverlay --media $VID_MEDIA \
  --prompt-text "chat, did i cook" \
  --params '{"fontFamily":"TikTok Sans SemiCondensed","position":"top-center","sizePercent":66,"fontSizeScale":0.5,"strokeWidth":4.5,"paddingTop":10}' \
  --wait -o with-text.mp4

Featured textOverlay + animatedCaptions presets. wonda edit {video,image,audio} accepts --preset <name> (scoped to --operation). --params fields override preset values on key collisions.

textOverlay (static, top-centered):

• TikTok White Highlight — black text on a slightly rounded white box.
• TikTok Black Highlight — white text on a slightly rounded black box.
• TikTok Red Highlight — white text on a slightly rounded red (#E14135) box.

animatedCaptions (STT-driven, bottom-centered):

• TikTok White Captions — black text, white highlight on the active word.
• TikTok Black Captions — white text, black highlight on the active word.
• TikTok Red Captions — white text, red (#E14135) highlight on the active word.

wonda edit video --operation textOverlay \
  --preset "TikTok Red Highlight" --media <id> \
  --params '{"text":"YOUR HEADLINE"}' --wait -o ./out.mp4

textOverlay renders locally via the bundled hyperframes (Chromium) renderer. There is no server-side image textOverlay anymore.

Font sizing guide:

• Static overlays: sizePercent: 66, fontSizeScale: 0.5, strokeWidth: 4.5
• Animated captions: sizePercent: 80, fontSizeScale: 0.8, strokeWidth: 2.5, highlightColor: rgb(252, 61, 61)
• Font: TikTok Sans SemiCondensed for both

Add animated captions (word-by-word with timing)

The animatedCaptions operation extracts audio, transcribes, and renders animated word-by-word captions — all in one step.

wonda edit video --operation animatedCaptions --media $VIDEO_MEDIA \
  --params '{"fontFamily":"TikTok Sans SemiCondensed","position":"bottom-center","sizePercent":80,"strokeWidth":2.5,"fontSizeScale":0.8,"highlightColor":"rgb(252, 61, 61)"}' \
  --wait -o with-captions.mp4

For quick static captions (no timing, just text on screen), use textOverlay with --prompt-text:

wonda edit video --operation textOverlay --media $VIDEO_MEDIA \
  --prompt-text "Summer Sale - 50% Off" \
  --params '{"fontFamily":"TikTok Sans SemiCondensed","position":"bottom-center","sizePercent":80}' \
  --wait -o captioned.mp4

Add background music

MUSIC_JOB=$(wonda generate music --model suno-music \
  --prompt "upbeat lo-fi hip hop, warm vinyl crackle" --wait --quiet)
MUSIC_MEDIA=$(wonda jobs get inference $MUSIC_JOB --jq '.outputs[0].media.mediaId')
wonda edit video --operation editAudio --media $VID_MEDIA --audio-media $MUSIC_MEDIA \
  --params '{"videoVolume":100,"audioVolume":30}' --wait -o with-music.mp4

Editor output chaining

When chaining multiple editor operations (e.g., editAudio → animatedCaptions → textOverlay), extract the media ID from each editor job output and pass it to the next step. Note the jq path differs from inference jobs:

# Inference jobs: .outputs[0].media.mediaId
# Editor jobs:    .outputs[0].mediaId

EDIT_JOB=$(wonda edit video --operation editAudio --media $VID --audio-media $AUDIO \
  --params '{"videoVolume":0,"audioVolume":100}' --wait --quiet)
STEP1_MEDIA=$(wonda jobs get editor $EDIT_JOB --jq '.outputs[0].mediaId')

CAP_JOB=$(wonda edit video --operation animatedCaptions --media $STEP1_MEDIA \
  --params '{"fontFamily":"TikTok Sans SemiCondensed","position":"bottom-center","sizePercent":80,"strokeWidth":2.5,"fontSizeScale":0.8,"highlightColor":"rgb(252, 61, 61)"}' --wait --quiet)
STEP2_MEDIA=$(wonda jobs get editor $CAP_JOB --jq '.outputs[0].mediaId')

wonda edit video --operation textOverlay --media $STEP2_MEDIA \
  --prompt-text "Hook text" --params '{"position":"top-center","fontFamily":"TikTok Sans SemiCondensed","sizePercent":66,"fontSizeScale":0.5,"strokeWidth":4.5}' --wait -o final.mp4

Merge multiple clips

Always merge locally with ffmpeg. Server-side merge (wonda edit video --operation merge) can hang for 30+ minutes once any input exceeds ~7MB.

Download every Wonda media ID, then concat. Stream-copy is fast but requires matching codec/profile/resolution; fall back to re-encode if it errors:

wonda media download $CLIP1 -o /tmp/clip-1.mp4
wonda media download $CLIP2 -o /tmp/clip-2.mp4
wonda media download $CLIP3 -o /tmp/clip-3.mp4
cat > /tmp/concat.txt <<EOF
file '/tmp/clip-1.mp4'
file '/tmp/clip-2.mp4'
file '/tmp/clip-3.mp4'
EOF
ffmpeg -y -f concat -safe 0 -i /tmp/concat.txt -c copy /tmp/merged.mp4
# If stream-copy fails, re-encode:
# ffmpeg -y -f concat -safe 0 -i /tmp/concat.txt \
#   -c:v libx264 -preset medium -crf 18 -pix_fmt yuv420p -movflags +faststart \
#   -c:a aac -b:a 192k /tmp/merged.mp4

# Re-upload only if a downstream wonda step needs the mediaId.
MERGED_MEDIA=$(wonda media upload /tmp/merged.mp4 --quiet)

File order in concat.txt = playback order. See the ffmpeg skill for the full concat reference.

Split scenes / keep a specific scene

Two modes, pick by intent:

# Split mode (default) — returns EVERY detected scene as its own media.
# JSON output lists each scene under scenes[] ({mediaId,index,startS,endS}).
wonda edit video --operation splitScenes --media $VID_MEDIA \
  --params '{"mode":"split","threshold":0.5,"minClipDuration":2}' --json
# With -o, each scene downloads to a numbered file (out-1.mp4, out-2.mp4, ...);
# a single detected scene writes the path verbatim.
wonda edit video --operation splitScenes --media $VID_MEDIA \
  --params '{"mode":"split","threshold":0.5,"minClipDuration":2}' -o scenes.mp4

# Remove a scene (omit mode) — removes one scene, merges the rest into one file.
wonda edit video --operation splitScenes --media $VID_MEDIA \
  --params '{"mode":"omit","threshold":0.5,"minClipDuration":2,"outputSelection":"first"}' \
  --wait -o without-first.mp4
# outputSelection (omit mode only): "first", "last", or a 1-indexed number = which scene to REMOVE

Use omit mode for "remove frozen first frame" (common with Sora videos). Use split mode to get all scenes as separate clips.

Image editing

Any image edit — img2img, background removal, crop, text overlay, vectorize — has its own skill with the full decision tree, aspect-ratio rules, and model waterfall for edits:

wonda skill get image-edit

One gotcha worth keeping here: image and video background removal use different models (birefnet-bg-removal vs bria-video-background-removal). Never swap them.

Lip sync (last-resort fallback — prefer native-audio video models)

Sora, Sora 2 Pro, Veo 3.1, Kling 3, and Seedance 2 all generate speech in any language with correctly synced mouth movements as part of the video itself. That path produces dramatically better results than sync-lipsync-v2-pro: better lip physics, better lighting, better costs, and no second inference round-trip. For any talking UGC, ad, or spokesperson video, put the dialogue directly in the video model's prompt — do not chain TTS + lipsync.

Only reach for sync-lipsync-v2-pro when the user EXPLICITLY supplies both a pre-existing video and a pre-existing audio clip and asks you to align the mouth to that audio. If a user asks for lipsync as the default method of making a character speak, push back: the native-audio video models are the better tool and work in any language.

wonda generate video --model sync-lipsync-v2-pro --attach $VIDEO_MEDIA,$AUDIO_MEDIA --wait -o synced.mp4

Video upscale

wonda generate video --model topaz-video-upscale --attach $VIDEO_MEDIA \
  --params '{"upscaleFactor":2}' --wait -o upscaled.mp4

Clipping (longform → vertical shorts)

wonda clipping takes a long video (podcast, interview, talking-head) and produces short vertical clips. Selection is LLM-driven and supports a natural-language --brief so you can ask for specific moments instead of generic virality.

V1 renders 9:16 with face-tracked reframe (LR-ASD active-speaker detection + One-Euro stabilizer, default) and the existing animatedCaptions op + a top-third hook overlay per clip. Pass --reframe blur-fill to keep the full landscape source inside a vertical canvas with a blurred background instead.

Translated captions: pass --caption-language <code> (ISO-639-1, e.g. en) to render captions in another language while keeping the original audio. Each clip's transcript is translated per sentence and the spoken timing is preserved, so captions stay in sync. Omit the flag to caption in the spoken language. On --restyle, the language is inherited from the parent job unless overridden, so you can cheaply spin out an English-captioned version of an already-clipped job (reuses the transcript, no re-transcription).

Async: POST /api/v1/clipping returns a clippingJobId; the CLI polls GET /api/v1/clipping/jobs/{id} under --wait. Pass --output <dir> and the CLI downloads each rendered clip + a plan.json.

Auth: included in the paid plan.

Source: --url accepts YouTube and direct mp4 URLs.

wonda clipping --url "<youtube-url>" --brief "the most controversial moments" --wait

YouTube links work; a long video can take several minutes to ingest before transcription starts. If a YouTube ingest fails, download the file locally and upload it first, then clip with --media:

yt-dlp -o /tmp/source.mp4 \
  -f "bv*[ext=mp4][height<=720]+ba[ext=m4a]/b[ext=mp4][height<=720]" \
  --merge-output-format mp4 "<youtube-url>"
MEDIA=$(wonda media upload /tmp/source.mp4 --no-transcode --quiet)

--no-transcode skips the server-side normalize so longform sources are usable in seconds instead of minutes (a 1 h upload otherwise transcodes for ~19 min before clipping can start). Clipping handles incompatible formats (AV1, rotated phone video) per selected clip automatically. Omit the flag for uploads you plan to publish or edit directly without clipping.

# Plan only — fast, no render
wonda clipping --media $MEDIA --brief "the most controversial moments" --dry-run --wait

# Full pipeline: select + render + download
wonda clipping --media $MEDIA \
  --brief "the most controversial moments" \
  --caption-preset "TikTok Red Captions" \
  --hook auto \
  --wait --output ./clips/

# Translate captions to English (original audio kept)
wonda clipping --media $MEDIA --caption-language en --wait --output ./clips/

# Re-caption an existing job into English (reuses STT + clip picks, ~$0)
wonda clipping --restyle <jobId> --caption-language en --wait --output ./clips/

# Filter by speaker (uses ElevenLabs diarization labels)
wonda clipping --media $MEDIA --speaker SPEAKER_00 --wait --output ./clips/

# Speaker rename for readable rationales
wonda clipping --media $MEDIA --speaker Joe \
  --speaker-map '{"SPEAKER_00":"Joe","SPEAKER_01":"Guest"}' --wait --output ./clips/

# Tune count and durations — pick a target length with a tolerance
wonda clipping --media $MEDIA --brief "punchy one-liners" \
  --count 5 --duration 20 --tolerance 5 --wait --output ./clips/

# Or specify an explicit min/max range instead (mutually exclusive
# with --duration/--tolerance)
wonda clipping --media $MEDIA --brief "punchy one-liners" \
  --count 5 --min-duration 8 --max-duration 30 --wait --output ./clips/

# Auto-pick FX preset per clip from a catalog
wonda clipping --media $MEDIA --auto-preset \
  --preset-catalog '[{"slug":"flash_glow","description":"glow + scene flash"},{"slug":"text_glow","description":"per-word text glow"}]' \
  --wait --output ./clips/

Job-status shape (returned by GET /api/v1/clipping/jobs/{id}):

{
  "clippingJobId": "...",
  "status": "succeeded",
  "stage": "succeeded",
  "progress": 1,
  "plan": {
    "sourceDurationSec": 1800.5,
    "speakers": ["SPEAKER_00", "SPEAKER_01"],
    "clips": [
      {
        "start": 12.4,
        "end": 38.7,
        "title": "Why he quit the agency",
        "hookText": "He admits…",
        "rationale": "Concedes \"the agency model is dead\" then explains why...",
        "score": 87,
        "dominantSpeaker": "SPEAKER_00",
        "reframeMode": "blur-fill",
        "preset": null,
        "mediaId": "uuid-of-rendered-clip",
        "url": "https://storage.googleapis.com/.../clip.mp4"
      }
    ]
  },
  "error": null
}

Editor operations reference

crop vs imageCrop: video crop is ratio/percent based (aspectRatio or cropPercent+cropAxis); it does NOT take pixel coordinates and rejects cropPixelX/Y/Width/Height with an error. For an exact pixel rectangle, use imageCrop. Run wonda operations info <operation> for the full param list, defaults, and ranges of any op.

Valid textOverlay fonts: Inter, Montserrat, Bebas Neue, Oswald, TikTok Sans, TikTok Sans Condensed, TikTok Sans SemiCondensed, TikTok Sans SemiExpanded, TikTok Sans Expanded, TikTok Sans ExtraExpanded, Nohemi, Poppins, Raleway, Anton, Comic Cat, Gavency Valid positions: top-left, top-center, top-right, center-left, center, center-right, bottom-left, bottom-center, bottom-right

Marketing & distribution

# Connected social accounts
wonda accounts instagram
wonda accounts tiktok

# Analytics
wonda analytics instagram
wonda analytics tiktok
wonda analytics meta-ads

# Organic owned-post analytics use platform-specific read commands:
# wonda x analytics, wonda linkedin analytics, wonda reddit analytics.
# The global analytics group above is for connected marketing APIs.

# Scrape competitors
wonda scrape social --handle @nike --platform instagram --wait
wonda scrape social-status <taskId>                   # Get results of a social scrape
wonda scrape cancel <taskId>                          # Cancel any public scrape task
wonda scrape ads --query "sneakers" --country US --wait
wonda scrape ads --query "sneakers" --country US --search-type keyword \
  --active-status active --sort-by impressions_desc --period last30d \
  --media-type video --max-results 50 --wait
wonda scrape ads-status <taskId>                      # Get results of an ads search

# Download a single reel or TikTok video
SCRAPE=$(wonda scrape video --url "https://www.instagram.com/reel/ABC123/" --wait --quiet)
# → returns scrape result with mediaId in the media array

# Publish
wonda publish instagram --media <id> --account <accountId> --caption "New drop"
wonda publish instagram --media <id> --account <accountId> --caption "..." --alt-text "..." --product IMAGE --share-to-feed
wonda publish instagram-carousel --media <id1>,<id2>,<id3> --account <accountId> --caption "..."
wonda tiktok creator-info --account <accountId>      # Live privacy options + comment/duet/stitch defaults
wonda publish tiktok --media <id> --account <accountId> --caption "New drop" --privacy PUBLIC_TO_EVERYONE
wonda publish tiktok --media <id> --account <accountId> --caption "..." --privacy PUBLIC_TO_EVERYONE \
  --disable-comment --commercial-disclose --brand-organic
wonda publish tiktok-carousel --media <id1>,<id2> --account <accountId> --caption "..." \
  --privacy PUBLIC_TO_EVERYONE --cover-index 0

# History
wonda publish history instagram --limit 10
wonda publish history tiktok --limit 10

# Browse media library
wonda media list --kind image --limit 20
wonda media info <mediaId>

X/Twitter

Supports reads, writes, and social graph.

⚠️ **Anti-frau

…