Sistava

HeyGen Video Translation: How to Dub With Lip Sync

How-to — — by Mahmoud Zalt

How to translate a video in HeyGen: upload rules, which engine to pick, how many languages per job, captions, credit costs and the duration limits.

You have a video that works in one language and an audience that speaks another. Video translation takes the original, produces a translated voice track, and in most modes redraws the speaker's mouth so the lips match the new words.

The mechanics are simple. The choices are where people lose time and credits, because the same feature is documented with different names on different HeyGen pages, and the engine you pick changes both the quality and the bill.

Start with one target language, watch the result end to end, and only then queue the rest. HeyGen's own help centre recommends exactly this, and it is the single habit that saves the most credits.

What can you actually upload?

MP4, MOV or WEBM files up to 5GB, or a public YouTube or Google Drive link. HeyGen's help centre documents three ways in: drag and drop into the upload area or click to upload, select an existing HeyGen project, or paste a URL.

One requirement catches people out repeatedly. Their documentation states the source video must be a single language. A recording that switches languages partway through is not a supported input, and the result will reflect that.

RequirementWhat is documented
File formatsMP4, MOV or WEBM
File sizeUp to 5GB
Duration range2 seconds to 10 hours, dependent on plan
LanguageThe source must be a single language
Link sourcesYouTube or Google Drive, publicly accessible
Per-video cap, Creator and Pro30 minutes
Per-video cap, Business60 minutes
Per-video cap, EnterpriseNo stated duration limit

Which translation engine should you choose?

Match the engine to how visible the speaker's face is. HeyGen's help centre documents three tiers, and the credit rate rises with how much visual work the engine has to do.

EngineCredit rateDocumented use
Audio Only2 credits per minuteWhen no lip sync is needed
Speed5 credits per minuteFront-facing video with minimal facial obstruction
Precision10 credits per minuteComplex angles, occlusions, multi-person conversations

If the speaker is offscreen, small in frame, or narrating over footage, audio only is both cheaper and better. There is nothing to lip sync, so paying five times the rate for lip sync buys you nothing.

How do you run the translation?

Open the translation tool, upload or link your video, set the source and target languages, pick an engine, then submit. HeyGen's Academy walkthrough describes reaching it from the dashboard by clicking Create and then Translate a Video.

Translating a video in HeyGen

  1. Check the source before you upload — One language throughout, clean audio, and within your plan's per-video duration cap. Their documentation lists 30 minutes on Creator and Pro and 60 minutes on Business, so a long webinar may need splitting first.
  2. Open the translation tool — HeyGen's Academy walkthrough describes clicking Create from the dashboard and then Translate a Video. Their help centre article describes reaching the Video Translation tool from your workspace.
  3. Add the video — Drag and drop the file or click to upload, choose an existing HeyGen project, or paste a YouTube or Google Drive link. Links have to be publicly accessible or the fetch will fail.
  4. Pick the engine that matches the shot — Audio only when the speaker is not clearly visible, the speed engine for straightforward front-facing footage, the precision engine when there are difficult angles, obstructions or several people talking.
  5. Set source and target languages — Choose the original language, then your targets. HeyGen documents both base languages and localised variants, using English with an American variant as their example, which matters when your audience expects a specific regional accent.
  6. Translate one language and watch it fully — HeyGen's own recommendation is to test one language before translating to all targets. Watch for names, product terms and numbers, which are where machine translation most often goes sideways.
  7. Decide on captions before you export — Their documentation notes that when captions are enabled you get two versions, one with captions burned into the video and one without, and that caption exports are .srt only. Decide up front, because burned-in captions cannot be removed later.
  8. Queue the rest and collect from your library — Once the test language reads well, run the remaining targets. HeyGen's Academy page says translated videos appear in your library and on the Share page, ready to review, download or share.

That is eight steps to move one video into one language, and most of them are checks rather than creative work. Multiply it across a content calendar and the checking becomes the job.

Repetitive checking is exactly the shape of work we took on at Sistava, in a different corner of the business. You hire an AI Employee and tell it in plain English which markets matter and what should go out where. It sets itself up from that brief and comes back to you when something needs a decision, rather than handing you another checklist to run.

Have someone who speaks the target language watch the test before you publish. An automated translation can be grammatically correct and still land badly, and the person most likely to notice is a native speaker, not the tool.

Whose voice and face are you translating?

If the person on screen is not you, get their permission first. Lip-synced translation puts new words in someone's mouth and moves their lips to match, which is a meaningful change to how they are represented, not a subtitle.

HeyGen's moderation policy requires consent from the person depicted for avatar creation and says removal requests must be honoured at any time. Whether or not a specific translation job triggers that exact rule, the underlying principle is worth applying yourself: do not publish a person saying something they did not say without asking them.

What does the translation not do for you?

It gives you a file. It does not decide which markets are worth entering, write the localised titles and descriptions, publish to the right channels on the right schedule, or answer the questions that arrive in a language you do not read.

That is the work a Sistava AI Employee picks up. You tell it in plain English that this video is going out in four languages and that you want the titles, descriptions and posts written for each, then the replies handled in the same language. It has the tools connected already and it asks you when it is unsure rather than guessing at your market.

HeyGen does the hard technical half of this genuinely well, and lip-synced dubbing that holds up on a front-facing shot is not a small achievement. The half that remains is coordination, and coordination is where most localisation efforts quietly stall.

FAQ

How do I translate a video with HeyGen?

Upload the video or paste a public YouTube or Google Drive link, choose the source language and your target languages, pick a translation engine, and submit. HeyGen's Academy walkthrough describes reaching the feature from the dashboard by clicking Create and then Translate a Video, with results appearing in your library and on the Share page.

How many languages can HeyGen translate a video into at once?

HeyGen's documentation gives different numbers on different pages. Their help centre article states up to 10 languages per job, while their Academy walkthrough describes a Quick Translate mode taking up to five target languages. Check what your own account offers, since this depends on the mode and plan you are using.

How many credits does HeyGen video translation cost?

Their help centre lists three engines with different rates: audio only at 2 credits per minute, the speed engine at 5 credits per minute, and the precision engine at 10 credits per minute. Their credits article separately describes video translation at 1 minute per 5 credits in standard mode with precision costing double. Confirm the current rate in your account before running a long batch.

Does HeyGen translation include lip sync?

It depends on the engine. The audio only option is documented as having no lip sync, which suits narration or offscreen speakers. The speed engine is documented for front-facing video with minimal facial obstruction, and the precision engine for complex angles, occlusions and multi-person conversations. Pick based on how clearly the speaker's face is visible.

How long a video can HeyGen translate?

Their help centre documents a supported duration range of 2 seconds to 10 hours depending on plan, with per-video caps of 30 minutes on Creator and Pro, 60 minutes on Business, and no stated duration limit on Enterprise. A long recording may need splitting before upload.

Can I get subtitles as a separate file?

HeyGen's documentation notes that when captions are enabled you receive two versions of the video, one with captions burned in and one without, and that caption exports are .srt only. Decide before generating, since captions burned into the picture cannot be stripped out afterwards.

Sources