Dubbed video: burn target-language captions with script_text

After you dub a video into Spanish or German, burn captions from the translation you already approved, not from a fresh transcription. Send the dubbed MP4 to Sume's video captions endpoint with your translated script as script_text. Speech-to-text still supplies the timings, and Sume aligns your wording onto them, so brand names and product terms appear exactly as written. The job is $0.20 for clips up to 60 seconds.

Why not let the transcriber write it

A transcriber hears the dubbed audio and writes what it heard, which may differ from your approved translation: a different spelling of a product name, a number written as digits or words, a dropped filler. If your translation went through review, that review is the source of truth. script_text keeps it that way.

The request

Set language to the dub language so recognition expects it. Name a style that fits the script: the Latin styles for Spanish or German, a Hangul-safe style for Korean.

curl -X POST https://api.sume.com/v1/video-captions \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: dub-es-caption-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/example/dub-es.mp4",
    "language": "es",
    "style": "slam",
    "script_text": "Conoce la nueva coleccion de verano."
  }'

When alignment fails

If the script and the speech differ too much, the job fails with script_alignment_mismatch or script_alignment_failed, and the suggested next step is to simplify the script or omit it. The usual cause in dubbing is that the dub speaks a slightly different sentence than the script, because the voice was given an edited version. Make the script you caption the same text you sent to text-to-speech, character for character, which should reduce alignment failures. The captions docs list the errors.

A note on voices and languages

Microsoft's MAI-Voice-2.1 page lists 23 languages (read 2026-10-04), and MAI-Transcribe-2-Streaming is described as covering 60 languages. Whatever produced your dub, the caption step here only needs a public HTTPS video URL, the language and the script. See the dubbing pipeline walkthrough for the steps before this one, and jobs and results for polling.