YouTube multi-language audio: add dubbed tracks, made with AI

YouTube's multi-language audio lets you add audio tracks in other languages to a single video or Short, and each viewer hears the track for their preferred language by default. You upload your own dubbed, audio-only tracks in YouTube Studio: the feature doesn't create them, and it's available to creators with access to Advanced features.

The YouTube rules below come from its help page Add Multi-language features to your videos, read on 2026-09-28. The steps for making a track with AI use the speech schemas in the Sume API reference and Sume's Timeline audio docs.

How does YouTube multi-language audio work?

One upload, several soundtracks. The rules that matter before you make any audio:

How do I add a multi-language audio track on YouTube?

YouTube's steps, in YouTube Studio on a computer:

• From the left menu, select Languages, then click the video.
• Click Add Language and pick the language.
• Next to Dub, click Add, then Select file, and choose the audio file.
• Click Publish. To replace a track later, click Delete under Audio for that language and upload again.

How do I make a dubbed audio track with AI?

Four steps per language: get the original words as text, translate them, speak the translation, and check the length.

• Text: use your script, or transcribe the original with POST /v1/stt-1.0/transcribe, which takes a public HTTPS audio_url. For a video already on Sume, POST /v1/audio-detach pulls its audio track out as wav or mp3 first.
• Translation: do it yourself or with an agent, and review it; YouTube recommends second-language tracks match the original as closely as possible.
• Speech: send the translated transcript to POST /v1/tts-1.0/generate with a voice and the track's language code in language (BCP-47 / ISO-639); omitted, it defaults to English. Multilingual text to speech API covers which languages to test and the voice-language check.
• Length: YouTube wants roughly the video's length. If the track runs long or short, set generation_config.speed between 0.6 and 1.5 and render again. For a script too long for one request, speak it in parts and join them with Timeline audio.

curl -X POST https://api.sume.com/v1/tts-1.0/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: yt-dub-es-001" \
  -d '{
    "transcript": "Hoy te enseño a configurar el panel en un minuto.",
    "avatar_handle": "acme",
    "language": "es",
    "generation_config": { "speed": 1.1 }
  }'

What does a dubbed track cost, and what are the limits?

Each step bills on its own, plus a 5.5% agent fee by default. TTS returns mp3 by default, or wav; confirm that the file type is one YouTube Studio accepts before you upload.

Is this the same as dubbing the video itself?

No. A multi-language audio track only swaps the sound; the picture, including the speaker's lips, stays the original. To re-render the video over a new voice, see Translate a video's voiceover, and Lip sync vs dubbing explains when matching lips matters.