You recorded one good video, and now you want it to speak Spanish, Hindi, German and Japanese without hiring four voice actors. That is the promise behind HeyGen Video Translate, a tool that swaps the original voice for a dubbed one and moves the speaker's lips to match. It sounds simple. In practice, people run into three walls: the bill is confusing, the API has its own rules, and jobs sometimes sit at "processing" forever. This article takes each wall in turn and ends with an online route you can use right now.
💡 Quick answer: Lip-synced translation is billed per minute of source video. Third-party sources put the API rate at roughly $2 per minute for lip sync and about $1 per minute for audio-only dubbing, while subscription plans count credits instead. Check HeyGen's own pricing page before you budget, because these numbers move.
What Labs HeyGen Video Translate Does
HeyGen Video Translate takes a finished video, transcribes the speech, translates the script, generates a new voice (often a clone of the original speaker), and re-renders the mouth area so the lips follow the new language. It started life in the experimental corner of the product, which is why you still see people searching for it as "HeyGen Labs". Over time it moved closer to the main dashboard, and some accounts now find it under the regular create menu. If you cannot find it where a tutorial says it should be, that is the first thing to check.
The Four Steps Under the Hood
- Transcription of the source audio into text.
- Translation of that text into the target language.
- Voice synthesis, usually keeping the speaker's tone and pace.
- Lip sync rendering, which edits the face so mouth shapes fit the new audio.
Each step can fail on its own, which explains why "not working" has so many different causes. A noisy recording breaks step one. A long idiom breaks step two. A face turned away from the camera breaks step four.

Who Actually Uses It
The usual crowd is easy to spot: YouTubers who want a second channel in another language, course creators, small companies with product demos, and marketers who need one spokesperson video to run in ten markets. The common thread is a talking-head clip where the face is visible and the audio is clean. That kind of footage is exactly where lip sync dubbing shines.
The Real Cost of Dubbing
Pricing is where most searches begin, and it is also where most blog posts disagree with each other. The pattern that shows up most often looks like this.
| Billing route | Lip sync translation | Audio-only dubbing | Notes |
|---|
| Subscription credits | About 5 credits per minute | About 2 credits per minute | Precision mode is reported at roughly double in some sources |
| API (pay as you go) | About $2.00 per minute | About $1.00 per minute | Billed from a separate wallet, not your plan credits |
| Per-video estimate for a 3 minute clip | About $6 | About $3 | Per target language, so multiply by the number of languages |
The Multiplication Problem
The number that surprises people is not the per-minute rate. It is the language count. A five minute video sent into six languages is thirty minutes of billable output. At the API figure above, that lands near $60 for lip sync, and it repeats every time you re-run a job after fixing a script error.
💡 Tip: Proofread the transcript before you start the translation. A misheard word in the source language produces a confidently wrong sentence in every target language, and you pay again to redo each one.
Credits Versus Wallet
Subscription credits and the API wallet are separate pools. People who buy a plan and then call the API are often confused when the API refuses to run even though the dashboard shows plenty of credits. The API draws from a USD balance that you top up on its own, and reports say you can start with a small amount to test.

Three Ways to Spend Less
- Trim before you upload. Cut intros, dead air and outros. You are billed on duration.
- Test on a 20 second clip. Check voice, accent and lip sync on a short sample before sending the full video.
- Use audio-only dubbing for B-roll heavy videos. If the speaker is rarely on screen, lip sync adds cost and nothing else.
How the API Works
Developers usually reach for the API once they have more than a handful of videos. The flow is the same across most video translation services: you submit a job, receive a job ID, and poll until the status says the output is ready. Then you download the file.
A Typical Request Flow
- Authenticate with an API key in a request header.
- Submit a public URL to your source video, the target language, and an optional mode flag for speed or precision.
- Store the job ID that comes back.
- Poll the status every 15 to 30 seconds, not every second.
- Download the result when the status changes to complete, and save it to your own storage because output links expire.
Limits Worth Planning Around
Third-party wrappers mention rate limits on the order of a hundred job creations per minute and about ten concurrent jobs per key. Treat those figures as a rough shape, not a promise, since the numbers can change. What matters for your code is the behavior: a 429 response means slow down, and a 5xx response means retry with an increasing delay.

Input Rules That Trip People Up
| Problem | What happens | Fix |
|---|
| Private or expiring video URL | Job fails right after submit | Use a public, stable link for the whole job |
| Very long source video | Job rejected or slow | Split into scenes under the length limit |
| Wrong file format | Immediate error | Export as standard MP4 |
| Multiple speakers talking over each other | Garbled voices and mismatched lips | Edit so one person speaks at a time |
Why It Stops Working
Now the part many readers came for. When Video Translate seems broken, the cause usually falls into one of five buckets.
Stuck on Processing
This is the most common complaint. A job sits at the same percentage for an hour. Before you do anything else, check the provider's status page and look at the specific Video Translate row, not just the overall banner. If that row shows a degraded state, waiting is the only real fix. If everything is green, cancel the job and submit it again.
Bad Transcript, Bad Translation
Open the transcript the moment it appears. Fix names, product terms and numbers. Brand names are a classic problem: "Kubernetes" becomes "cooper nettys" in the source text, then gets translated into nonsense.
Lip Sync Looks Off
The mouth movement drifts or looks rubbery when the speaker turns their head, covers their mouth with a hand, or wears a mask. Frames with a clear, front-facing view of the face work best. Fast cuts between camera angles can also confuse the face tracking.

Credits or Wallet Errors
If the dashboard says you have credits but the API says you have none, you are probably looking at two separate balances. Top up the API wallet or switch the job to the web interface.
Language Mismatch
Choosing the wrong source language is a silent killer. The tool transcribes a Portuguese clip as if it were Spanish, and everything downstream is wrong. Set the source language by hand when you can, instead of relying on auto detection for accented or mixed-language speech.
💡 Troubleshooting order: status page, then transcript, then source audio quality, then face visibility, then billing balance. Most failed jobs get solved by the second step.
Online Options That Work Today
You do not have to commit to a subscription or write API code to dub a video. There are online tools where you upload an MP4, choose a language, and download the result. On PicassoIA you can run the same HeyGen engine directly in the browser, which is useful when you want to test one clip before deciding whether a bigger setup is worth it.

Dub With Lip Sync
The HeyGen Video Translate page offers lip-synced dubbing in more than 150 output languages and regional variants, including options like Arabic (Egypt), Spanish (Mexico), Chinese (Cantonese, Traditional) and English (Australia). It accepts standard MP4 files and gives you a choice between a speed mode and a precision mode.
Pick a Lip Sync Engine Directly
If you already have the dubbed audio and only need the mouth movement, two sibling models handle that job. Lipsync Precision favors quality, and Lipsync Speed favors turnaround. They suit workflows where you write the translation yourself or use a human translator.
Translate Audio Without Touching the Face
When the speaker is off camera, or when you do not want any facial edits, ElevenLabs Dubbing translates the speech into dozens of languages while keeping the original video untouched. That is often cheaper and faster for screen recordings, interviews with many cutaways, and podcasts with static visuals.
Build Your Own Voiceover
Some creators prefer to translate the script by hand and generate the voice separately. ElevenLabs v2 Multilingual produces voiceovers in 30 or more languages, and Realtime TTS 2 gives you natural-language control over delivery. Pair either with a lip sync model above and you have full control over every sentence.

How to Use HeyGen Video Translate on PicassoIA
Here is the short path from a single MP4 to a dubbed version.
Step 1: Prepare the Source Clip
Export your video as MP4. Keep the speaker's face visible and well lit, and make sure only one person talks at a time. If the clip is long, cut it down to the part you want translated first.
Step 2: Open the Model
Go to the HeyGen Video Translate page on PicassoIA and upload your MP4 in the Video field.
Step 3: Choose the Output Language
Open the output_language list. It defaults to English, and it includes broad choices like Spanish, French, Hindi and Japanese, plus regional variants such as French (Canada), Portuguese (Brazil), English (UK) and Spanish (Mexico). Pick the variant your audience actually speaks, since word choice and accent change from region to region.
Step 4: Set the Mode
The mode setting has two values. Speed is built for quick turnaround and works well for drafts and social clips. Precision is the default and aims for higher quality output, which is the better pick for anything customer-facing.

Step 5: Run and Review
Start the job and watch the result with the sound on. Check three things: the voice sounds like a plausible version of the speaker, names and numbers are correct, and the lips match during close-ups. If something is off, fix the source or switch the language variant and run again.
💡 Parameter tip: Run a 15 to 20 second test in speed mode first. Once the voice and accent feel right, rerun the full video in precision mode.
A Simple Test Plan
| Test | What to look for |
|---|
| 20 second clip, speed mode | Accent, tone, pacing |
| Same clip, precision mode | Lip sync quality, voice clarity |
| Full video, one language | Transcript accuracy, timing across scenes |
| Full video, all languages | Consistency and brand name handling |
Getting Better Results Every Time
Good output starts before the upload. These habits fix most quality complaints.
Record With Dubbing in Mind
- Speak at a steady pace with small pauses between sentences. Translated phrases often run longer, and pauses give them room.
- Avoid slang and long jokes that depend on wordplay.
- Keep your face in frame and avoid covering your mouth.
- Use a decent microphone. Clean audio gives the transcript fewer chances to go wrong.
Plan for Text Length
Languages differ in how many words a sentence needs. German and Spanish tend to expand, while Chinese and Japanese often compress. If your video has fast on-screen callouts, leave extra time around them so the dubbed speech does not rush.
Keep a Review Checklist
- Names, brands and numbers spelled correctly in the transcript.
- Regional variant matches the target audience.
- Lip sync checked on close-ups and on the first and last five seconds.
- Subtitles added for viewers who watch on mute.
- Final file saved to your own storage.

When a Synthetic Presenter Makes More Sense
If you do not have a source video at all, or your goal is to publish many localized versions of a script, a talking presenter built from text can save the filming step. Avatar V creates talking avatar videos, and P Video Avatar is another way to animate a speaker from a photo and audio. These are different tools from a dubbing workflow, so pick them when you are starting from a script instead of footage.
Is It Worth Paying For
The answer depends on volume and on how visible the speaker is.
| Situation | Verdict |
|---|
| One or two videos per month, a few languages | Use a pay-per-run online tool and skip the subscription |
| Weekly videos in many languages | An API or plan starts to pay off, but test costs first |
| Mostly screen recordings or voiceover | Audio dubbing is enough, skip the lip sync |
| Customer-facing brand video | Use precision mode and have a native speaker review it |
A native speaker review matters more than people expect. Machine translation is good, but a human check catches the odd phrase that sounds stiff, the unintended double meaning, or the term your industry uses differently. Even a ten minute pass from a colleague is cheap protection for a video that carries your name.

Try Your Own Dub on PicassoIA
The fastest way to settle the pricing and quality questions is to run a short test on your own footage. Upload a 20 second clip to HeyGen Video Translate, pick a language, compare speed and precision modes, and watch where the lips and voice land. If you only need the voice, try ElevenLabs Dubbing. If you want a presenter from a script, try Avatar V. Every model is listed at picassoia.com/en/all-models, so you can mix dubbing, voice, lip sync and image tools in one place and see which combination fits your channel. Pick one clip, pick one new language, and press run.