A client once sent me a two-minute product video in English and asked a simple question: can this sell in Germany too? The video was good, but almost nobody in his target market was going to watch a sales video in a language they do not speak, and subtitles alone were not closing the gap. Viewers read subtitles. They trust a voice that speaks to them.
AI video dubbing with lip sync solves exactly that problem. It translates the spoken words, revoices them in the new language, and reshapes the speaker's mouth so the video looks like it was recorded in German, Spanish or French from the start. I deliver this as a paid service, and this guide is my exact workflow: transcript first, translate for the ear, pick the voice tool for voice quality and the lip-sync tool for the picture, and never ship without a full human watch-through in the target language.
If you have no one to film at all, an AI spokesperson video can present your script for you.
Quick Verdict:
- Dub with lip sync when a visible speaker talks to the camera and the video already works in one market.
- Use a voiceover dub without lip sync when the video is mostly product shots, screen recordings or B-roll. It costs less, and nobody sees a mouth to mismatch.
- Use subtitles only when the budget is near zero or the video is a quick internal update. Subtitles inform, but they do not sell the way a native voice does.
- Wait if the video itself still needs work. Dubbing multiplies a proven video. It does not rescue one nobody watched in its original language.
Contents:
- Voice quality and lip sync are two different problems
- Step 1: Get the transcript right before anything is translated
- Step 2: Translate, then proofread for the ear, not just the eye
- Step 3: Render the voice, where voice quality wins
- Step 4: The lip-sync pass, where the picture catches up
- Step 5: The manual fixes that make it client-ready
- What this costs a small business, honestly
- Who should dub, and who should wait
- Frequently Asked Questions
Voice quality and lip sync are two different problems
This is the mistake most people make when they pick a dubbing tool. They find one tool that sounds beautiful and assume the lips will match, or one tool whose lips match and assume the voice will sound natural. In my experience, different tools win each problem, and a professional result usually means using each tool for what it is best at, or accepting one deliberate trade-off and telling the client about it upfront.
Keep that split in mind through every step below. It is the reason the workflow has separate passes instead of one button press. The table below is the decision I make before any job starts.
| Option | What changes | Use it when |
|---|---|---|
| Dubbing with lip sync | Voice and mouth movement both match the new language | A visible speaker talks to the camera |
| Voiceover dub | Voice changes, picture stays as shot | Product shots, screen recordings, B-roll |
| Subtitles only | Text on screen, original voice stays | Near-zero budget, internal updates |
Step 1: Get the transcript right before anything is translated
I start with the original video's exact words, not a rough summary. If the client has a script, I check it against the audio line by line. Scripts and final edits often differ by a word here and there, and a dub built on the wrong script will drift out of sync with the picture in ways that are hard to fix later.
If there is no script, I transcribe the audio and read it back while watching the video once. This pass also catches the things a translator needs to know: names that must stay unchanged, numbers, and any joke or idiom that will not survive a literal translation.
Step 2: Translate, then proofread for the ear, not just the eye
A translation that reads well on paper can still sound wrong spoken aloud. German sentences run longer than English ones. Spanish moves faster. French has its own rhythm. A line that takes four seconds in English might take six in German, and if the speaker on screen stops moving their mouth at four seconds, the illusion breaks.
So the proofread is done out loud. I read each translated line against the original clip's timing and trim or rephrase where a line overruns its shot. This is a normal part of the job, not a failure, and it is the step most one-click dubbing skips, which is why so many AI dubs feel slightly off even when the words are correct.
Step 3: Render the voice, where voice quality wins
For the voice itself, I work with tools built for voice quality. ElevenLabs Dubbing (version 2) is the one I reach for when the voice has to carry the video. According to ElevenLabs' own documentation, Dubbing v2 translates video into 90 or more languages while keeping the character of the original speaker, and it is built to preserve emotion, delivery and timing. Its honest limit, which I tell clients upfront, is that the dubbing pipeline has no lip sync: the voice will sound right, but the mouth will not automatically match. Free-plan dubs are watermarked and limited to about 10 minutes a month for non-commercial use, so client work runs on a paid plan.
The choice at this step is deliberate: if the video is a talking-head ad where the face is on screen the whole time, voice alone is not enough and we go to Step 4. If it is a voiceover-led video, product shots, screen recordings or B-roll, a great voice without lip sync is often the right, cheaper answer. Voice quality matters beyond dubbing too: the same care about how an AI voice sounds to a real caller is what I test when I build an AI voice agent for a dental clinic.
Step 4: The lip-sync pass, where the picture catches up
When the face matters, the dubbed audio goes through a lip-sync pass. HeyGen Video Translate is the tool I use here: it dubs with lip sync built in, so the speaker's mouth is reshaped to match the new language. HeyGen's own product page lists translation into 175 or more languages with voice cloning and lip sync. HeyGen's help centre prices Video Translate by engine (figures reported 30 October 2026): audio-only translation at 4 credits a minute, Speed lip sync at 6 credits a minute, and Precision lip sync at 10 credits a minute. Credits are shared across all HeyGen features, so a long video eats the same budget as other HeyGen work, which matters when quoting a job. The free plan covers 3 videos a month, watermarked, which is fine for testing but not for delivery.
Lip sync is not magic, and I check every pass for its three usual failures: mouth shapes that lag the audio by a beat, teeth and mouth edges that smear on fast speech, and profile or half-turned shots where the model has less face to work with. Those shots get one of three fixes: a shorter rephrased line, a cutaway to B-roll over the worst half-second, or, honestly, sometimes, the client accepting a voiceover-style version for that section.
Step 5: The manual fixes that make it client-ready
No AI dub ships without a human watch-through, start to finish, in the target language. I check four things on this pass:
- Sync drift. Does the audio stay matched from the first line to the last, or does it slowly slide? Drift means a timing fix, not a re-render.
- Names and numbers. Product names, prices and phone numbers survive translation only if someone checks them. I check every one.
- Breathing room. A dubbed line that runs right up to the next cut feels rushed. A half-second of space at the end of a line is worth more than a perfect translation.
- One language, one listen. I listen to the whole video in German (or Spanish, or French) without the English in my head, the way the end viewer will. Anything that sounds translated rather than spoken gets reworked.
Only then is the video exported, in the same resolution and format as the original, with the original file kept untouched as the master.
If the source clip is blurry, sharpen it first: AI video upscaling can improve low-resolution footage so the lip sync still looks clean.
Once the dubbed file is exported, finish it with CapCut auto captions and clean audio before you post it.
What this costs a small business, honestly
The honest answer is that it depends on length, language and how much of the video is a talking head versus voiceover. The two cost drivers are the per-minute tool credits and the human proofread-and-fix time, and the second one is usually bigger than the first. A one-minute voiceover dub and a five-minute talking-head dub in three languages are very different jobs, and anyone quoting a single flat price for both is guessing.
For HeyGen Video Translate, the credit cost per finished minute depends on the engine you pick:
| HeyGen engine | Credits per minute | What it does |
|---|---|---|
| Audio only | 4 | New voice track, lips unchanged |
| Speed lip sync | 6 | Lip sync for a front-facing speaker |
| Precision lip sync | 10 | Lip sync through speaker switches and side profiles |
On the Creator plan at $29 a month with 600 credits, those rates buy roughly 150 minutes of audio-only dubbing, 100 minutes of Speed lip sync, or 60 minutes of Precision lip sync, before the human time is counted. Budget per language, not per video: every new language is a fresh render with its own credits.
What I can say plainly: AI dubbing has made a job that used to need a studio, a voice actor and a separate editor per language into something a small business can actually afford, provided the person doing it is honest about the lip-sync limits and does the manual pass. Skip that pass and you save money and lose the sale, because viewers can feel a bad dub even when they cannot explain why. Prices and limits in this guide were checked 31 October 2026. Tools change pricing often, so confirm on the pricing page before paying.
Who should dub, and who should wait
Dub when you already have a video that works in one market and a real audience in another language: a product that sells, a course with sign-ups, a service page with traffic. Dubbing multiplies a proven video. It does not rescue a video nobody watched in its original language.
Wait if the video itself still needs work, or if you are guessing at the market. Fix the message first in one language, prove it, then translate the winner. And if the video is made for social media, settle the platform question before you pay for versions: read Instagram vs TikTok for small business: which one to choose first, then dub for the platform that fits.
Once the dubbed versions exist, they still need a publishing rhythm, or they sit unwatched in a folder. Use the same system I teach to plan 30 days of social media content in one afternoon, and give each language version its own slot in the month.
Frequently Asked Questions
Do I need lip sync, or is a voiceover dub enough?
If a visible speaker talks to the camera, you need lip sync, or the mismatch will distract every viewer. If the video is product shots, screen recordings or B-roll, a voiceover dub is enough and costs less, because nobody sees a mouth to mismatch.
How much does AI video dubbing cost?
Tool credits are only part of the bill. HeyGen Video Translate runs at 4 credits a minute for audio-only, 6 for Speed lip sync and 10 for Precision lip sync, and the Creator plan is $29 a month with 600 credits. The human proofread, timing fixes and final watch-through usually cost more than the credits, and every extra language is a fresh render, so budget per language.
Can AI dubbing clone my own voice into another language?
Yes. Both tools in this workflow offer voice cloning, so the dubbed version can keep the character of the original speaker instead of swapping in a generic narrator. Only clone a voice you own or have clear permission to use. A clone without permission is not a shortcut, it is a problem.
Will viewers know the video was dubbed?
A good dub feels native, and most viewers simply watch it. A bad dub also gets felt, even when viewers cannot explain why: the timing feels off, the mouth lags, or a line sounds translated instead of spoken. The manual watch-through in Step 5 is what moves a dub from the second feeling to the first.
Which languages work best for AI dubbing?
German, Spanish and French are the languages I deliver, and they work well because the voice models for them are strong. Widely spoken languages with strong voice models are the safest first step. For a rarer language, check the tool's language list and test a short clip before promising a full video.
Written by Amjad Rasool, SEO Expert and Social Media Expert. I help small businesses grow with content systems, local SEO and AI tools. About Amjad Rasool
Thinking about dubbing a video for a new market? Ask me in the comments which video you would translate first. I read every question.
){kind=link}
0 Comments