ElevenLabs dubbing and voice conversion

Use this skill when the task is about transforming existing speech: translating a source audio/video into another language, converting a performed voice into a selected voice, cleaning speech before conversion, or planning localization deliverables around ElevenLabs. Do not use it for ordinary text-to-speech generation unless the TTS output is only one supporting step in a dubbing or voice-conversion workflow.

This skill is evidence-aware. Treat dated vendor facts below as documented facts verified on 2026-07-10. Treat operational observations and production heuristics as guidance to test on the current project, not as provider guarantees.

Choose the ElevenLabs path

Use the decision before touching files:

• Dubbing/localization: choose ElevenLabs Dubbing when the source is an audio or video asset that must be translated into another language while preserving speaker timing, delivery, and identity as much as the product supports. Plan for transcript review, language QA, and media export.
• Voice changer / speech-to-speech: choose the Voice Changer API when the language/content is already correct and the job is to transform an existing performance into another available or cloned voice while retaining emotion, rhythm, timing, and delivery.
• Voice isolation: choose Audio Isolation before dubbing or voice conversion when noise, music, ambience, reverb-like background, or production sound makes speech recognition/conversion unreliable. Do not use it blindly on clean audio; extra processing can remove wanted texture or introduce artifacts.
• External localization: if exact legal/medical/brand translation, rigorous subtitle timing, or lip-sync under broadcast constraints matters, plan a human translation/review layer around ElevenLabs rather than presenting an automatic dub as final.
• Generic TTS: use a different ElevenLabs TTS skill for scripted voiceover from text. This skill only covers TTS-adjacent decisions where existing speech, speakers, timing, localization, or conversion are central.

Documented facts verified 2026-07-10

ElevenLabs publishes an existing Dubbing API at POST /v1/dubbing that accepts multipart source media or a source_url, target_lang, optional source_lang, optional num_speakers, time range, watermark and background-audio controls, and automatic/manual mode. The response includes dubbing_id and expected_duration_sec. GET /v1/dubbing/{dubbing_id} returns status, source language, target languages, created time, editability, media metadata, and failure error text. GET /v1/dubbing/{dubbing_id}/audio/{language_code} streams the dubbed MP3/MP4 result for the original automatic dub; if a dub was edited in Dubbing Studio, use the resource render workflow rather than the original automatic-audio endpoint. Transcript export is available at /v1/dubbing/{dubbing_id}/transcripts/{language_code}/format/{format_type} with srt, webvtt, or json, and source may be used for the original media language. Sources: https://elevenlabs.io/docs/api-reference/dubbing/create, https://elevenlabs.io/docs/api-reference/dubbing/get, https://elevenlabs.io/docs/api-reference/dubbing/audio/get, https://elevenlabs.io/docs/api-reference/dubbing/transcripts/get.