Quick Answer
Translating a video has six practical problems, and each has a known fix. The voice: dubbing tools now keep the speaker's own voice. The mouth: close ups need lip sync, which HeyGen Video Translate adds and ElevenLabs Dubbing does not. The price: on Percify a dub is 72 credits a minute, about $1.52 at the Creator rate, or 180 a minute with lip sync. Names and product terms: read the translated script before rendering. Text burned into the picture stays in the original language. And a dubbed voice is realistic AI speech, which TikTok and the EU AI Act require you to disclose.
Six real problems with translating a video, from keeping the voice and fixing lip sync to pricing, mistranslated terms and disclosure, each with the fix and what it costs. Checked 15 September 2026.
Related next steps
Most articles about video translation list vague "challenges" and then claim one tool fixes every one of them. The useful version names each problem, the fix that actually exists, and what it costs. Prices below were read on each tool's own page on 15 September 2026; we make Percify, so read its lines with that in mind.
This is one of three pages: for the process, see translating a video of yourself, step by step; for choosing a method, dub, subtitle or re record.
1. The voice stops sounding like the speaker
Old dubbing replaced the speaker with a voice actor. Current tools keep the original voice: ElevenLabs delivers dubs in a clone of the speaker, Rask clones the speaker's voice in 32 languages, and both of Percify's dubbing models keep the voice. Fix: choose a tool that dubs in the speaker's own voice, and check that it does before paying.
2. The mouth moves in the wrong language
A dub changes the audio, not the face, so on a close up the lips still form the original words. Fix: lip sync. HeyGen Video Translate ↗ resyncs the lips as part of the translation, at 3 credits a second on Percify. ElevenLabs Dubbing ↗ lists no lip sync because it is an audio dub, which is fine for wide shots, voiceovers and screen recordings. For a close up dubbed without it, you can run the dubbed audio through InfiniteTalk Fast ↗ at 2 credits a second.
3. The price is hard to predict
Tools price translation per minute, per plan pool or per credit, which makes them hard to compare. On Percify, ElevenLabs Dubbing is 72 credits a minute, about $1.52 at the Creator rate, and HeyGen Video Translate is 180 credits a minute, about $3.80. Rask worked out to $1.32 a minute on its Creator pool and $0.78 on Creator Pro, and ElevenLabs charged $2.23 to $2.70 for extra minutes. Fix: price one real video at the length you actually make, into one language, since every language is its own run. The full breakdown is in how much AI dubbing costs.
4. Names, prices and product terms come out wrong
Machine translation is weakest exactly where mistakes are most visible: brand names, prices, and technical words. Fix: read the translated script before you render, not after, and fix those terms by hand. It is the cheapest step in the whole process and the one that prevents a second paid run.
5. Text in the picture stays in the original language
Titles, captions and labels burned into the video are pixels, not audio, so a dub leaves them untouched. Fix: redo on screen text separately, and remake thumbnails in the target language before you publish.
6. A dubbed voice is AI speech, and has to be disclosed
A clone of someone's voice speaking a language they never recorded is realistic AI generated speech. TikTok requires a label ↗ on realistic AI content including AI speech, and Article 50 of the EU AI Act ↗, which has applied since 2 August 2026, requires deepfake audio resembling a real person to be disclosed. Fix: label the translated version, and keep the speaker's written consent if the voice is not yours.
Where to start
For a short clip, HeyGen's free plan translates 3 videos a month of up to 3 minutes each. For your own videos at volume, Percify runs both models from one balance, with plans from $26 a month and credits refunded when a run fails. The whole process is in the AI video dubbing guide, and plans are on pricing.
Ready to Create Your Own AI Avatar?
Make a talking video of yourself from one photo, in your own voice. No watermark on any plan.
Cancel anytime, no contracts
Got questions?
Frequently asked
Six practical ones: keeping the speaker's voice, the mouth still moving in the original language on close ups, unpredictable pricing, mistranslated names and product terms, text burned into the picture staying untranslated, and the duty to disclose a dubbed voice as AI speech.
On current tools, yes. ElevenLabs dubs in a clone of the original speaker, Rask clones the speaker's voice in 32 languages, and both of Percify's dubbing models, ElevenLabs Dubbing and HeyGen Video Translate, keep the voice.
On Percify, read on 15 September 2026: ElevenLabs Dubbing is 72 credits a minute, about $1.52 at the Creator rate, and HeyGen Video Translate with lip sync is 3 credits a second, 180 a minute, about $3.80. Rask worked out to $0.78 to $1.32 a minute on its yearly pools, and ElevenLabs charged $2.23 to $2.70 for extra minutes. Each language is a separate run.
Use a translation that resyncs the lips, such as HeyGen Video Translate, which does it on every translation, or run the dubbed audio through a lip sync model such as InfiniteTalk Fast. An audio only dub leaves the mouth forming the original words, which shows on close ups.
Usually yes. TikTok requires a label on realistic AI generated content including AI speech, and EU AI Act Article 50, applying since 2 August 2026, requires deepfake audio that resembles a real person to be disclosed.
