HeyGen Video Translation: Engines, Credits and Steps

HeyGen's video translation dubs an existing video into another language, clones the speaker's voice by default and, on two of its three engines, re-renders the lips to match the new words. You pay in credits per minute of source video, per target language: 4 for Audio only, 6 for Speed and 10 for Precision.^[1] Paid plans reach 175+ languages and dialects, with a per-video cap of 30 minutes on Creator and Pro or 60 minutes on Business.^[2] One job takes up to 10 target languages.^[3] Our verdict: Speed covers most talking-head videos. Proofreading needs Pro. Free is only a test.

On the Free plan you get a limited trial of lip-sync translation, capped at 1 minute per video and 30+ languages, and its videos are watermarked. Its output is licensed for personal, non-commercial use only, so a translated sales demo, client video or ad needs a paid plan from day one. Treat Free as a way to hear your own cloned voice in another language and judge the lip sync on your footage, not as a way to ship anything.

Pick the engine before you upload

Audio only dubs the voice and leaves the picture alone, with no lip sync. It fits narration over slides, screen recordings and podcasts, where nobody watches a mouth. Speed adds lip sync tuned for front-facing videos; HeyGen positions it for quick drafts and faces with little movement. Precision gives the highest-quality lip sync and copes better with speaker switches, side profiles and hands or objects in front of the face. It re-renders mouth movements with several models, so it takes longer to process.

Your footage matters as much as the engine. HeyGen's help warns that lip sync may not be perfect when a speaker's head turns more than 30 degrees from the camera or the face is too small in the frame. Its developer guidance adds that stable, well-lit shots at 720p or higher with few cuts work best, while poor audio and fast cuts degrade results. The source video must be in one language. Mixed-language interviews need splitting first.

The price gap is easy to see. A 10-minute video dubbed into three languages is 30 billable minutes. That costs 120 credits on Audio only, 180 on Speed and 300 on Precision, by our arithmetic. Precision is two-thirds dearer than Speed for every minute. Use it where an angled or busy face needs it, or for the final brand version. Run drafts on Speed.

What each plan lets you translate

HeyGen's pricing page lists Creator at $29 a month billed monthly or $24 a month billed yearly ($288 a year) with 600 credits a month, Pro from $49 a month billed monthly or $41 a month billed yearly ($488 a year) for 1,000 credits, and Business at $149 a month billed monthly or $119 a month billed yearly ($1,428 a year) for the first seat with 1,500 credits shared by the workspace. Credits, not the plan price, decide how many minutes you can dub, and monthly credits roll over only one extra cycle, as our HeyGen pricing guide explains.

In translation terms, HeyGen's help puts those allowances at about 100 Speed minutes on Creator, 166 on Pro's entry tier and 250 on Business. By our arithmetic, the same credits give 60, 100 and 150 minutes on Precision. Long videos burn credits fast. One 30-minute webinar dubbed into 10 languages is 300 billable minutes: 1,800 credits on Speed or 3,000 on Precision. That is beyond both Creator's 600 credits and Pro's entry tier, and more credits never raise the per-video length cap.

The choice between the two cheapest paid plans comes down to volume and review. Their credits cost almost the same, so by our arithmetic a Speed minute per language works out to about 29 cents of plan price on either at monthly billing. Moving up one plan costs $20 a month more at monthly billing and adds 400 credits, Proofread and 4K output. Below about 100 Speed minutes a month with no script review, stay on Creator. Above that, Creator cannot buy top-ups on the web, so step up.

Step by step: from upload to dubbed video

1. Open Video Translation and upload an MP4, MOV or WebM file up to 5 GB, between 2 seconds and 5 hours long, or paste a YouTube or Google Drive link. Vimeo links are not accepted.
2. Set the source language, which must be a single language for the whole video.
3. Pick up to 10 target languages.
4. Choose Audio only, Speed or Precision.
5. Open the Advanced tab to pick a brand glossary and set voice enhancement and Dynamic Duration.
6. Turn captions on now if you want them.
7. Submit, or on Pro and above, open Proofread first.

Dynamic Duration is on by default. It lets each segment stretch or shrink by up to 20 percent so the translated speech fits, which can change the total length of the video. Billing still follows the source duration. Captions are worth deciding early: switching them on produces two outputs, one with burned-in captions and one clean, and caption files export only as SRT. Without proofreading, captions must be added before you submit, and style, color and speed controls are not available.

HeyGen's help estimates processing at about 5 minutes per minute of content, proofreading included; that is render time, and it says the regular queue averages just under 24 hours on weekdays. Creator and Pro process three videos at once and Business six, so extra jobs wait as Pending. If the source language is unsupported, the job fails instantly and costs nothing. If moderation blocks a translation, you can ask for a manual review by emailing HeyGen's moderation team with the video ID.

Proofread and the glossary decide the quality

Proofread, also called Review & Edit, is available on Pro, Business and Enterprise.^[4] It lets you check the translated script, voices and timing before the final render. The Preview tab plays the translated audio over the original picture without lip sync. The Result tab appears only after you click Generate Result. You can export the script as Excel or SRT, edit it offline and re-upload it. Version history restores earlier drafts, and a Stale badge flags script changes since the last render.

Two timing rules catch people. If you already have your own translated SRT, attach it before creating the Proofread project, or the segment counts may not match. And once a translation is fully generated it can no longer be proofread. Fix the words first, then spend the Precision credits. HeyGen's guidance recommends a proofread pass for videos over 3 minutes and for corporate or high-stakes content.

The brand glossary keeps product names and fixed terms consistent. Force Translate sets an exact translation for a term, and Don't Translate locks words such as brand names; entries go in by hand or by CSV upload. If one term carries both rules, Don't Translate wins. Pronunciation rules change only the spoken audio, so captions keep the original spelling. Enterprise adds proofreader invitations by email (they need only a free HeyGen account, take no seat and see one language), preset stock voices instead of cloning the speaker, and custom SRT input in Precision mode.

What the dub leaves behind

Only spoken audio is translated. On-screen text and hard-coded captions stay in the original language, and sound effects are always removed from the translated audio track. A training video with labelled diagrams, or a product ad built on a music bed with effects, will need rework after dubbing. For videos made in HeyGen itself, translating from the share page creates a draft in the target language, though timing and length may shift by language.

Delivery is where the feature earns its keep. Videos grouped in a collection get a multilingual player with a language dropdown that can be embedded on any page or in an LMS. Viewers can toggle captions with the CC button on a translated video's share page. Business and Enterprise add SCORM and xAPI export of translations, which suits course teams; our guide to HeyGen for education covers that workflow.

Translating in bulk: batches and the API

Batch Translation in the app collects every output, plus the source video, in an automatic Collection folder. Through the API, one batch call takes up to 100 items after language expansion, so a video sent into five languages counts as five items. Items process independently, which means one bad source does not fail the rest. Batch endpoints are limited to 30 calls a minute on every plan.

The billing line matters here. API calls draw on a separate prepaid wallet, not on your plan credits, and so do the translate actions in Zapier and Make. The Remote MCP server, used from assistants such as Claude or Cursor over OAuth, is different: its translation tool spends your plan credits. Our HeyGen API guide walks through the wallet and the limits.

The bottom line

HeyGen's translation is a solid way to dub talking-head and training videos into many languages from one upload, provided you budget in credits per minute per language rather than in plan price. Start on Speed, film front-on, keep text off the screen, and use Proofread before any Precision render. Creator handles light volume without review; Pro is the sensible floor once scripts need checking or monthly minutes pass about 100. Many languages need a bigger Pro tier; videos over 30 minutes need Business (60 minutes), whose 1,500 credits may need $5-per-100 top-ups.