How to Dub a Video With AI: Step-by-Step Guide (2026)

To dub a video with AI, upload a finished video you own, pick the target language, and let the tool transcribe the speech, translate it and re-voice it, ideally in the original speaker's own cloned voice. Then check the translation, decide whether you need lip sync or captions, and export one version per language. On Kenerate's AI Dubbing tool that takes one upload and costs 3 credits per second of the source file (as of 24 October 2026).

This guide walks through the whole process step by step: how to prepare the source, which settings matter, when voice cloning beats a stock voice, when lip sync is worth paying for, what it costs and the mistakes that make AI dubs sound fake.

What does "AI dubbing" actually do?

AI dubbing chains four jobs that used to need a translator, a voice actor and a sound engineer:

1. Transcription: speech recognition turns the dialogue into a timestamped script and works out who is speaking.
2. Translation: the script is translated into the target language, line by line.
3. Voice generation: the translated lines are spoken again, either in a cloned copy of each original speaker's voice or in a stock voice.
4. Mixing: the new speech is laid back onto the video, usually over the original music and ambience.

It is easy to mix dubbing up with things that sound similar, so here is how they differ:

If you only need text, add subtitles to the video instead. It is cheaper and faster. If you need people to hear your video in their language, you need a dub.

What do you need before you start?

• A finished edit. Lock the picture first. If you trim a shot after dubbing, the timing of every line after it breaks.
• Clean dialogue. Speech recognition is the first link in the chain, so errors made there carry through to the translation and the voice. Loud music under the voices, wind noise and people talking over each other all lower the quality.
• Rights to the footage and the voices. Only dub content you made or have permission to use. Cloning someone's voice into another language is still using their voice. Get their consent, and don't dub other people's videos or celebrities.
• A short glossary. Write down product names, people's names and jargon that must not be translated. You'll use it when you review.

How do you dub a video with AI, step by step?

Here is the workflow on Kenerate. The same logic applies to most dubbing tools.

Step 1: Upload the video or audio file

Open AI Dubbing in the app (Tools → AI Dubbing) and upload your file. Kenerate accepts a video up to 500 MB or an audio file up to 50 MB, up to 15 minutes long. A video comes back as a video with the new audio track. An audio file (a podcast, a voice note) comes back as audio.

Longer than 15 minutes? Split it at natural scene breaks and dub each part separately. That also makes fixing one part cheaper.

Step 2: Choose the source and target language

Leave the source language on Auto-detect unless detection gets it wrong. Kenerate recognises 57 spoken source languages. Then pick the target: there are 33, including English, Hindi, Spanish, French, German, Portuguese, Japanese, Korean, Chinese, Arabic, Tamil and Indonesian. The source and target can't be the same language.

One dub produces one language. If you want three languages, you run three dubs.

Step 3: Set the number of speakers

Leave speakers on auto (0) and the tool will detect up to 32 voices. If you know exactly how many people talk, say an interview with two people, set it to 2. This helps the tool keep each person's lines in that person's voice instead of mixing them up.

Step 4: Decide what happens to background audio

By default the original music and ambience are kept, and only the voices are replaced. Turn on drop background audio if the bed is noisy or the music fights with the new voices. You get cleaner speech, but the soundtrack is gone, so you'd need to add music back in an editor.

Step 5: Check the price and run it

The price is the length of the file: 3 credits per second, rounded up to whole seconds.

For scale, the $15 Starter plan includes 1,300 credits, which covers a little over 7 minutes of dubbing. There's only one engine and no cheaper quality setting. If a dub fails at the provider, the credits are refunded automatically.

Step 6: Review the dub like a viewer would

Watch the result at normal speed, with sound, before you publish. Check:

• Names, numbers and claims. These are the errors that cause real damage. A price, a date or a dosage translated wrongly is worse than an awkward phrase.
• Your glossary terms. Dubbing engines sometimes translate brand names or jargon loosely.
• Pacing. Some languages take more words than others to say the same thing, so a line can feel rushed when the dub has to fit the original timing.
• Voice consistency. Each speaker should sound like the same person all the way through.

If you have a native speaker in your audience or team, a five-minute review from them is worth more than any automatic score.

Step 7: Add captions or lip sync if needed, then export

Two optional finishing steps, both covered below: captions in the new language, and lip sync for on-camera speakers. Then download and publish one version per language.

Should you use voice cloning or a stock voice?

Kenerate's dubbing re-voices each speaker in their own voice, so your presenter still sounds like your presenter in Spanish or Japanese. ElevenLabs, whose dubbing model powers the feature, says its dubbing keeps each speaker's emotion, timing and tone (ElevenLabs docs).

If the original speaker is your brand, clone. If nobody on screen is talking to camera, a well-chosen stock voice reading a translated script often sounds just as good. You also get full control over the script.

Do you need lip sync for a dubbed video?

Dubbing changes the audio, not the picture. Kenerate's dubbing doesn't lip sync: mouths keep their original movement. Whether that matters depends on what's on screen:

• No lip sync needed: voice-over on B-roll, screen recordings, slides, product shots, wide shots, animation without visible mouths, podcasts.
• Lip sync worth it: close-up talking heads, ads with a presenter facing the camera, testimonials.

For those cases, dub first, then run the dubbed video through a video-to-video lip-sync model in the AI lip sync generator. Prices as of 24 October 2026:

Test a 10-second close-up first. Lip sync is the step where cost and quality vary the most.

How do you add captions in the new language?

Dubbing doesn't add on-screen text. For captions that match the new audio, run the dubbed video through auto subtitles. Don't use the original-language captions: captions made from an earlier script won't match what viewers now hear. Kenerate's subtitle tool handles clips up to 4 minutes, at 20 credits for up to a minute in 720p (doubled at 1080p), with 10 caption styles.

How do you publish dubbed versions on YouTube?

You have two options:

1. One video, several audio tracks. In YouTube Studio on a computer, open Languages, pick the video, click Add language, then Add next to "Dub" and upload your dubbed audio. YouTube asks for an audio-only file of roughly the same length as the video, and this feature needs access to Advanced features (YouTube Help). Export the audio track from your dubbed file for this.
2. Separate videos or channels per language. This is simpler, and it works on TikTok, Instagram and anywhere else that has no audio-track feature.

YouTube also has its own automatic dubbing for eligible creators. It isn't available for videos over 120 minutes or with little speech, the language pairs are limited (Chinese, Japanese and Thai originals can only be dubbed into English), and you can't edit the automatic dubs (YouTube Help). If you want to choose the languages, keep your own voice everywhere or review the script, dub it yourself and upload the track.

What are the most common AI dubbing mistakes?

Which way of dubbing should you choose?

For most creators the sensible default is AI dubbing with cloning, a native-speaker check, captions on the dubbed file, and lip sync only where faces fill the frame. If you're translating many short clips into English, the video translator page shows the same workflow from the translation side.

Bottom line

Dubbing a video with AI is now a one-upload job: file in, language picked, dub out in the speaker's own voice. The quality comes from what you do around it. Start from a clean, locked edit, set the speaker count, review names and numbers, caption the dubbed file, and lip sync only where faces fill the frame. Try one 30-second clip first (90 credits on Kenerate), check it with a native speaker, then dub the rest.

Updated 24 October 2026. Prices and limits are from Kenerate's live tool configuration on that date and may change.