A dub is not a voiceover. A voiceover sits on top of a video that was already finished in one language. A dub has to make the same performance work in another language — the mouth, the timing, the emotion, the pauses — without the viewer feeling the seam.
That used to mean a studio, a cast and a mixing session. Today the whole chain fits in a browser tab: transcribe the speech, translate it, generate a voice, and re-render the lips so they follow the new audio. This guide walks through what each step actually does, the rules that separate a watchable dub from an uncanny one, and how to do it end to end.
What AI dubbing actually does
Dubbing is a pipeline, not a single model. Each stage can fail on its own, which is why a dub that sounds fine in your head can fall apart on screen.
| Stage | What happens | What to check |
|---|---|---|
| Transcribe | Speech is turned into text, with timestamps | Names, numbers and jargon |
| Translate | Each line is rewritten in the target language | Meaning and tone, not word-for-word |
| Voice | A voice performs the translated lines | Emotion and pacing |
| Sync | The lips are re-rendered to match the new audio | Mouth shapes on stressed syllables |
| Assemble | The new audio is laid back onto the video | Music, pauses and room tone |
Subtitles stop after the translation stage. A dub goes all the way to the voice and the lips. That extra distance is what makes a dub feel native — and what makes the source footage matter so much.
The rules that decide whether a dub works
Start with clean speech. The transcription stage hears what your microphone heard. Wind, traffic, echo and heavy room noise all become translation errors before you ever see them. A quiet room beats an expensive microphone in a loud one.
Use one speaker per shot. The pipeline maps one voice to one mouth. Overlapping speakers, background chatter and duets make that mapping ambiguous, and the mouth starts moving on the wrong syllables.
Keep the face visible. A dub can only work with a mouth it can see. Front-on, well-lit faces with an uncovered mouth sync reliably. Profiles, hands over the mouth, masks and fast cuts are where the uncanny valley lives.
Translate for the time you have. A line that takes four seconds in the original has to land in roughly four seconds in the new language. A literal translation that runs long gets compressed by the voice — or forces the edit to cut words that mattered. Short sentences are your friend.
Match the emotion, not just the words. The same sentence can be a joke, a warning or a sales pitch. If the voice reads it flat, the dub is technically correct and emotionally wrong.
How to dub a video with AI, step by step
Step 1: Prepare the clip. Use MP4 or MOV, 720p or 1080p, between 2 and 60 seconds, up to 100 MB. Trim to the part that matters — billing is per second, and a tighter clip is easier to sync. One person facing the camera gives the cleanest result. If the source is a TikTok video, you can paste the link instead of downloading it first.
Step 2: Pick the language and the voice. Upload the clip or paste the TikTok link, choose one of the 40 target languages, and pick the voice the dub should speak with. The source language is detected automatically, so you don't have to label it. The price is shown before anything runs.
Step 3: Run it and compare. Every result plays side by side with the original in a before/after slider. Watch the mouth on stressed syllables first — that's where a bad sync shows up before anything else. Listen for the pauses, too: stretches without speech keep the original audio, so intros, music breaks and silences should still feel like the same video.
Step 4: Download and review. If the dub lands, download the MP4. If it doesn't, check the source before re-running: a front-on shot, a single speaker and clean audio solve more failed dubs than any setting.
The run is billed per second of video at 10 credits per second, so a thirty-second clip costs 300 credits. The exact price is always shown before you submit, and failed runs refund automatically. There is no free tier for translation.
What AI dubbing is good at — and what it isn't
It is very good at the work that used to need a studio: turning a product clip into five language versions, localizing short-form content without a cast, making a founder video travel, or re-voicing archive footage for a new audience.
It is not a magic wand for every video. A noisy street interview, a shot with three people talking over each other and a two-minute monologue all fight the pipeline in different ways. The fix is usually editorial: cut to the clean shot, split the long clip, or record a tighter script in the target language.
One honest limitation worth knowing before you start: music playing under speech is not preserved in this version. The voice replaces the original audio where someone speaks; pauses and stretches without speech keep the original sound. If the music is the point, plan around it — or leave those sections untouched.
Rights and disclosure
Only dub footage you have the rights to use. A dub can make someone appear to say something they never said, so don't use a real person's face or voice to misrepresent them. When the dubbed version could be mistaken for the original, say that it's dubbed. That's not just a legal hedge; it's how you keep trust with the audience you're translating for.
FAQ
Is AI dubbing the same as subtitles? No. Subtitles keep the original audio and add text. A dub replaces the spoken audio in the target language and, in our translator, re-renders the lips to match. Subtitles are faster; a dub is what makes a clip feel native.
How long can the video be? From 2 to 60 seconds per run. For longer material, split it into segments and dub the part that carries the message. Billing is per second, so shorter runs are cheaper too.
Which languages are supported? Forty target languages, including English, Chinese, Japanese, German, Spanish, French, Korean, Portuguese and Arabic, with a choice of voice for each. The source language is detected automatically. Videos already in the target language are declined.
What does it cost? 10 credits per second of video — a thirty-second clip is 300 credits. The exact price is shown before you submit, failed runs refund automatically, and there is no free tier.
Does it keep background music? Stretches without speech keep the original audio and picture. Where someone speaks, the new voice replaces the original audio entirely; background music under speech is not preserved in this version.
Can I dub a YouTube video? You can upload any file you have the rights to use. TikTok links can be pasted directly; other platforms need the file downloaded first.
Does it change the person's face? No. The face, expression, lighting and camera stay as they were — only the lip region is re-rendered to follow the new voice.
Ready to hear your clip in another language? Dub your video — upload a clip or paste a TikTok link, pick a language, and compare the result side by side.