Dubbing Agent: voice track + subtitles in 18 languages

What it does

Takes a video and a translation script, and returns the video with a new voice track speaking that language - plus subtitles.

What it actually does, precisely. It synthesises the supplied script in the target language with a neural voice, muxes it as the first audio track, and keeps the original audio as the second track so nothing is lost. It also emits an SRT whose timings are distributed across the spoken duration.

What it does not do. It does not transcribe the original speech and it does not machine-translate. Send the translation you want spoken. This is stated up front because a dubbing tool that silently guesses at the dialogue produces confident nonsense.

Input

{ "input": {
    "video_url": "https://example.com/ad.mp4",
    "target_lang": "ar",
    "script": "النص الذي تريد أن يُنطق" } }

target_lang takes an ISO code: en, ar, fr, de, es, it, pt, ru, tr, hi, ja, ko, zh, nl, pl, id, sv, uk.

What you get back

• The dubbed MP4 (original audio preserved as track 2)
• An SRT in the target language
• A note of the voice track length

FAQ

Will the voice match the original speaker? It is a neural voice of the target language, not a clone of the original speaker. Cloning is not offered.

Can it handle a 40-minute lecture? The script cap is 12,000 characters per job, roughly 15-20 minutes of speech. Split longer material across jobs.

Does it lip-sync? No. Audio is replaced, the picture is untouched.