/pika:language-swap

Tools below are Pika MCP tools, named bare — call each under whatever prefix your session exposes for the Pika MCP.

Translate and dub a video into another language while preserving the original speaker's voice. Pipeline: dub (one worker call) → lipsync (default ON) → burn target-language captions or bilingual captions.

The dubbing worker does the heavy lifting in a single call: it transcribes, translates, preserves each speaker's voice server-side (no separate clone step), and returns a fully A/V-synced video — so there is no manual transcribe/clone/TTS/replace chain to manage and no duration-drift handling to do by hand.

Segmented / multi-language dub (per-range languages)

Use this when the user wants different languages on different parts of one video (e.g. first half Spanish, second half Japanese), or wants to translate only some sections and keep the rest in the original voice. Both are the same thing: a timeline of segments, each tagged with a language; any uncovered range keeps the original audio.

dubvideo takes a segments plan instead of targetlanguage (pass exactly one — they are mutually exclusive):

dub_video(source_video_url=<video_url>, segments=[
  {start_s: 0,  end_s: 30, target_language: "es"},
  {start_s: 30, end_s: 60, target_language: "ja"}
])

How to build the plan: the user needs to know where the content is before they can pick ranges, so transcribe first — extract the audio with extractaudiofromvideo, then transcribeaudio(audio=<audiourl>, timestamps=true), show the user the timestamped segments, and let them say which time range goes to which language. Then assemble segments[] (seconds, ordered, non-overlapping) and make ONE dubvideo call. There is no separate "video understanding" tool — the timestamped transcript is the understanding step.

Behavior of the segmented path:

• Shared voice across all segments. The source speaker is cloned once and every segment — in every language — is spoken in that same cloned voice, then the clone is recycled, all inside the one dub_video call. You never clone or delete a voice yourself.
• Keep-original. Any time range NOT covered by a segment plays the original audio (voice + background) untouched. To translate only parts of a video, list only the parts you want translated.
• Length-locked. Output stays exactly the source length (each dubbed range is speed-fit to its window), so boundaries line up with the original timeline.
• Provider. Mixed-language-per-range always uses the voice-cloning route automatically; the single-call whole-video dubbing route can't mix languages per range, so don't force a single-call provider for a segmented plan. Every covered language must be supported on the voice-cloning route — if one isn't, surface the error and consult references/language-coverage.md.
• Result. Same dubbed-video result; targetlanguage echoes the covered languages comma-joined (e.g. "spa,jpn"), and no single transcriptlanguage is returned (the track is multi-language). Lipsync (Step 2, default ON, ≤5 min) still runs on the whole dubbed video. For captions (Step 3), use the returned multi-language subtitles[] in caption_mode="manual"; auto re-transcription can't pick a single language for a mixed track.

If dub_video rejects segments (older deployment without segmented support), fall back to dubbing each range single-language and concatenating — but prefer the one-call segmented path when available.

Behavior defaults

• Target language: required via --to <language>. Prefer language codes: es, fr, ja, de, pt-BR, zh-Hans. The dubbing worker accepts ISO/BCP-47-like tags and normalizes script/region subtags before calling ElevenLabs (for example zh-Hans → zh; zh-Hant-TW → zh).
• Lipsync: ON by default — re-matches the speaker's mouth to the translated audio (fal sync-lipsync; the full-video lip-matcher, distinct from the portrait-image animator). Pass --no-lipsync to skip it when the source has no on-camera face or to avoid the meaningful cost (~$4/min on the sync-2-pro tier). Applies only to videos ≤5 min — edit_lipsync hard-caps at 300 s upstream, so longer sources auto-skip lipsync (see Step 2); the dub itself has no length limit.
• BGM / background music: kept by default — the dub lays the translated voice over the original music / SFX bed. Pass --no-bgm for a translate-only output: the worker drops the original music and keeps only the translated speech (dropbackgroundaudio=true).
• Captions: target-language captions are burned by default. When the user asks for bilingual / dual subtitles, burn the target-language (translated) row on top and the source-language (original) row below it — after dubbing, the translated speech is what's actually being said, so it's the primary row; the original is the secondary reference.
• Bilingual captions: enable when the user passes --bilingual-subtitles or asks for "bilingual subtitles", "dual subtitles", "two-language captions", "original + translated subtitles", "双语字幕", or "原文+译文字幕".
• Language coverage: if language support is questioned or a language-related upstream error occurs, consult references/language-coverage.md. Do not proactively surface provider-specific language-list details in normal user replies.

State variables produced and consumed

• video_url: input — from positional arg
• sourceinputurl: original positional URL — preserved for diagnostics if video_url is rehosted
• target_language: text — from --to <language>
• with_lipsync: boolean — defaults true; false only when --no-lipsync
• nobgm: boolean — true when --no-bgm (maps to dropbackground_audio=true)
• bilingual_subtitles: boolean — true when the user asks for bilingual / dual subtitles
• dubbedvideourl: dubbed, A/V-synced video — produced by Step 1
• dub_subtitles: optional target-language timed subtitles from the dub result — consumed by Step 3
• source_subtitles: optional source-language timed subtitles from the dub result — consumed by Step 3 for bilingual captions
• dubtranscriptsrt: optional target-language SRT from the dub result — returned for review/debugging
• sourcetranscriptsrt: optional source-language SRT from the dub result — returned for review/debugging
• sourcetranscriptlanguage: optional source-language code from the dub result
• lipsyncedvideourl: dubbed video with mouth re-matched — produced by Step 2 (when lipsync runs)
• captiontargetvideo_url: final visual video URL before captions are burned
• finalvideourl: video with target-language captions burned in — produced by Step 3

Step 0 — Parse input

Required:

• Positional video_url — MUST be https://...
• --to <language> — target language (free-text or BCP-47 code)

Optional:

• --no-lipsync — skip the default mouth-matching step.
• --no-bgm — translate-only output; drop the original music/SFX bed.
• --bilingual-subtitles — burn source-language + target-language subtitle rows.

Infer bilingual_subtitles=true from user wording even if the explicit flag is absent.

If --to is missing, STOP and prompt the user — UNLESS the user wants different languages on different parts, or to translate only some sections: that is the per-range segmented path (see "Segmented / multi-language dub" above), which uses a segments plan instead of --to.

For the segmented path, first build the time-range plan: extract the audio with extractaudiofromvideo, then transcribe it with timestamps via transcribeaudio(audio=<audio_url>, timestamps=true), show the user the timestamped segments, and capture which time range maps to which language into segments[].

Outputs: videourl, targetlanguage, withlipsync (default true), nobgm (default false), bilingual_subtitles (default false).

Step 1 — Dub the video (state: dubbedvideourl)

Call dub_video with:

• sourcevideourl — <video_url>
• targetlanguage — <targetlanguage> (ISO/BCP-47-like tag, e.g. es, pt-BR, zh-Hans)
• source_language — "auto"
• dropbackgroundaudio — true only when no_bgm is set; otherwise omit (keeps the original music bed)

dub_video is a plugin-only Pika MCP tool: the claude.ai connector surface can lag the plugin build, so a connector session may not expose it at all. If it is absent from your tool list, say so rather than substituting another tool.

dubvideo is worker-backed: if the response comes back as {taskid, status}, poll taskstatus until completed, then read the dubbed video from the result (videourl for a video source; audiourl for an audio source). Also capture optional subtitles[], transcriptsrt, and transcript_language — these are target-language transcript metadata the dub worker produced, consumed in Step 3.

For bilingual captions, also capture optional sourcesubtitles[], sourcetranscriptsrt, and sourcetranscript_language. These source-language transcript fields are best-effort. The dubbed media is still valid when transcript fields are absent.

Source not worker-fetchable: if dub_video fails because the source URL cannot be fetched — especially HTTP 403 / 4xx, hotlink protection, UA-gated hosts (Wikimedia/news CDNs), or "Access Denied" errors — do not keep retrying the same call. Rehost first:

1. Download the source bytes in the client/host environment using a normal browser/download path or an HTTP client with a real user-agent.
2. Call upload_asset with the downloaded filename, MIME type, and exact byte size, then upload the bytes to the returned presigned URL.
3. Set sourceinputurl = <original URL> and replace videourl with the returned Pika CDN publicurl. Do not construct CDN URLs manually.
4. Retry Step 1 once against the Pika CDN URL. All later steps must use the updated video_url.

If the client/host also cannot download the source bytes, stop and tell the user the host blocks direct fetch; ask them to upload the file or provide a different URL.

Outputs: dubbedvideourl, dubsubtitles, sourcesubtitles, dubtranscriptsrt, sourcetranscriptsrt, sourcetranscriptlanguage.

Step 2 — Lipsync (state: lipsyncedvideourl)

Default ON. Skip entirely when --no-lipsync is passed (then Step 3 captions dubbedvideourl directly).

Hard 5-minute cap — check duration before calling. editlipsync enforces a 300-second (5-minute) audio limit upstream (sync.so) and rejects anything longer with invalidinput before billing; every variant tier shares the same cap, so falling back through tiers does NOT help. If the dubbed video's durationseconds (returned by Step 1) is > 300, skip lipsync entirely, go straight to Step 3 captioning dubbedvideourl, and tell the user lipsync isn't available past 5 minutes (the dub itself works at any length). Only run the lipsync call below when durationseconds ≤ 300.

Cost heads-up first. Lipsync is the dominant cost (~$4/min on the v2-pro tier). Before calling it, estimate from the dubbed video's durationseconds (returned by Step 1) using costusd = durationseconds / 60 * 4. Do not round short clips up to a full minute: a 7s clip is about $0.47, not about $4. Send the user a one-line heads-up with the actual duration and prorated estimate, e.g. "Lipsync on — ~7s video, est. ~$0.47 at $4/min (pass --no-lipsync to skip). Starting now." If durationseconds is missing and no reliable media duration is available, state the $4/min rate without inventing a dollar estimate. Then proceed straight into the call; this is a heads-up, not an approval gate.

Call editlipsync(videourl=<dubbedvideourl>) with no audio_url — the worker syncs to the dubbed video's own embedded translated audio. Do not extract the audio just to feed it back in. (variant defaults to v2-pro, with sync-3 / v2 as fallbacks.)

Outputs: lipsyncedvideourl (read from url of response). When this step runs, Step 3 captions this video, not dubbedvideourl — otherwise the lip-matching is dropped.

Step 3 — Burn target-language captions (state: finalvideourl)

Caption the final video so the output carries readable subtitles (matches the common "translate + subtitle" expectation). Set captiontargetvideourl to lipsyncedvideourl when lipsync ran (the default), or dubbedvideo_url when --no-lipsync skipped it.

If this request is part of a Double video / split-screen comparison flow, build that Double video first and set captiontargetvideo_url to the final composed video URL. Do not burn captions onto only one panel before the Double video is composed; the bilingual caption burn should happen once, on the final visual output.

Call addcaptions once on captiontargetvideourl.

When bilingualsubtitles=true, use manual bilingual mode if both tracks are available: call addcaptions(videourl=<captiontargetvideourl>, captionmode="manual", subtitles=<dubsubtitles>, secondarysubtitles=<sourcesubtitles>, language=<targetlanguage>, secondarylanguage=<sourcetranscriptlanguage if available>, secondarysubtitlesposition="below", style="branded-space-mono", position="bottom"). The target-language (translated) row is the primary subtitles and renders on top; the source-language (original) row is the secondary reference and renders below it (secondarysubtitlesposition="below") — after dubbing the translated speech is what's actually spoken, so it leads. It works for every dub worker provider branch as long as dub_video returns both subtitle tracks.

If bilingualsubtitles=true but sourcesubtitles is missing, fall back to target-language captions only and tell the user the source transcript was unavailable from the dubbing provider. Do not invent a source-language row by retranscribing the final dubbed audio; that audio is already in the target language.

For target-language-only captions, prefer the target-language subtitles the dub worker already returned: if dubsubtitles is non-empty, call addcaptions(videourl=<captiontargetvideourl>, captionmode="manual", subtitles=<dubsubtitles>, language=<target_language>, style="branded-space-mono", position="bottom"). Manual mode skips a duplicate transcription pass and preserves the dubbing provider's target-language text.

If dubsubtitles is missing, empty, or rejected by addcaptions, fall back to auto: call addcaptions(videourl=<captiontargetvideourl>, captionmode="auto", language=<target_language>, style="branded-space-mono", position="bottom"). Auto mode re-transcribes the dubbed audio; use it only as the fallback because it costs extra time and can introduce CJK/proper-noun drift.

Use style="branded-space-mono" unless the user asks for a punchier style (tiktok / hormozi / karaoke). Skip this step only if the user explicitly asked for audio-only dubbing with no captions.

Outputs: finalvideourl (read from url of response).

Step 4 — Return

Reply with finalvideourl + the translated transcript (from dubtranscriptsrt / the dub result) for user review.

Offer a bilingual-subtitle version. When this run burned target-language-only captions (bilingualsubtitles=false) and a source transcript is available (sourcesubtitles is non-empty), close the reply by asking whether the user also wants a dual-subtitle version, e.g. "Want a bilingual version with the original + translated subtitles stacked? I can add it." If they say yes, re-run Step 3 in bilingual manual mode on the same captiontargetvideourl (the pre-caption visual video) — no re-dub or re-lipsync is needed, only the caption burn changes — and return the new finalvideourl. Skip the offer when bilingual captions were already burned (bilingualsubtitles=true), or when source_subtitles is missing — without a source transcript a bilingual version can't be produced (the dubbed audio is already in the target language), so do not offer what can't be delivered.

Failure modes