Can ChatGPT Translate and Dub a Video? What It Does and What It Hands Off
ChatGPT translates what is said in a video and will write you a timed subtitle file. It does not keep the speaker's voice, re-time the lips, or return an MP4 — those are three separate jobs, and this is what each one needs.
No — not end to end. ChatGPT can transcribe the speech in a video, translate it well, adapt the register, and hand you timed subtitles. It cannot speak that translation in the original speaker's voice, move their mouth to match the new language, or give you back a rendered video file. Dubbing is those last three things. Translation is the part it is genuinely good at, and also the smallest part of the job.
What ChatGPT actually does well here
Give it the audio or a transcript and it will produce the editorial half of a dub, which is the half most tools do badly:
- a clean transcript with speaker turns
- a translation that reads like speech rather than like a document
- a script cut to the original timing — the hardest editorial constraint in dubbing, because the dubbed line has to end when the original line ends
- flagged idioms, jokes and brand terms that should not go literal
- an SRT or VTT file, if subtitles are all you needed
If subtitles are an acceptable answer to your problem, you are already finished and it cost nothing. Most people asking this question want a spoken dub, so read on.
The three jobs it hands off
1. Voice preservation. A dub in a stock narrator voice over your own face reads as a re-upload, not as your video. Keeping your timbre means cloning it from the source audio — inference over a waveform, not text generation.
2. Lip-sync. The mouth in the existing frames says the old words. Matching the new ones means re-generating the mouth region frame by frame against the new audio.
3. Render and mux. New audio aligned back to the original timeline, encoded to an MP4 you can post. A chat window returns text and images; it does not return an encoded video.
What the dubbing step costs
fal-ai/heygen/v2/translate/precision runs all three in one call at $0.05 per second — $3.00 per minute, checked 10 August 2026. Inputs up to 8 minutes, 175+ target languages. The precision and speed variants are priced identically, so there is no reason to run speed.
| Job | Who does it | What it costs |
|---|---|---|
| Transcript, translation, timing-fit script | ChatGPT | included in whatever you already pay |
| Subtitles only | ChatGPT | same |
| Voice + lip-sync + rendered MP4 | a dubbing model | ~$3.00 per minute |
Our own revoice tool is that one call behind an upload form — you pick the language, you get the vertical MP4 back. The chat half of the workflow still belongs in chat.
How to check the result before publishing
Watch the first and last thirty seconds; sync drifts at the end of long footage, not at the start. Read the translated transcript for idioms that went literal. Check any close-up where lips fill the frame — reported sync accuracy sits in the 85–95% band, which reads as slight softness in a wide shot and as a mismatch in a close-up. Language-specific traps are worth knowing in advance: Russian expands against the original timing, and Spanish forces a regional choice before you start.
The part no tool asks about
Translating your own video is straightforward. Translating someone else's is a licensing question, and the obligation is yours whether the assistant, the dubbing model or the front end ever raises it. Get the rights first.