How to Translate a Video With AI in 2026

To translate a video with AI, transcribe the speech, translate the text, then publish subtitles or a dubbed voice track. Choose the output first, because it sets the price. HeyGen's Creator plan starts at $29 a month and its translations include lip sync, while Kapwing's free plan translates up to 5 minutes but watermarks every export. The prices below were read on each vendor's pricing page in October 2026.

Toolradar data: the AI voice ranking evaluated 125 AI voice tools, updated October 2026. This guide does not rank them. It shows what each output costs, using four tools and a free transcription and translation route.

Pick the output before the tool

Three outputs cost very different amounts. Subtitles are the lightest: the viewer reads the translated text while the original audio plays. A voice-over replaces the speech with a synthetic track over the original picture. Lip sync goes further and moves the speaker's mouth to match the new language, which matters on a talking-head video and costs the most.

If the video is a screen recording or a tutorial, subtitles usually do the job. If the viewer will listen rather than read, choose a voice-over. Pay for lip sync only when the face fills much of the frame, because a mouth that does not match the words is the first thing a viewer notices. For the wider field of dubbing tools, see our best AI dubbing tools list, which ranks eight of them.

Step 1: Get a transcript with timestamps

Every route starts with a transcript that has timestamps, because the timestamps decide where each subtitle or dubbed line sits. You have two options.

The first is the transcript the tool builds from your upload. Upload the file, then read the source text the tool produces before you move on. Fixing a misheard word here costs seconds. Fixing it after the translation costs a re-run.

The second is Whisper, OpenAI's open-source speech model. Its code and model weights are released under the MIT License, so you can download and run it yourself. You need ffmpeg installed first. Use Whisper when you want a copy of the transcript that sits outside any vendor, or when you plan to translate in a different tool.

Whisper's translate option has a limit that catches people out. According to the Whisper README, the translate task translates into English only. The README also says the turbo model returns the original language even when translate is set. For a Spanish or German subtitle file, transcribe in the source language first, then translate the text in Step 2.

Step 2: Translate the transcript, not the video

Translate the transcript file. Text is cheap to fix, and the timestamps stay where they are, so a corrected word does not throw off the timing. DeepL's free plan gives each user 50,000 characters of text a month, plus one document a month up to 5 MB. DeepL shows its paid prices in euros only, so this guide quotes no paid DeepL figure. For the paid tiers, see our DeepL pricing breakdown, and for how DeepL compares with Google's translator, read DeepL vs Google Translate. If you need a translation engine for documents and apps rather than one transcript, the best AI translation tools list compares the options.

Then check the translation against the source before you use it. Machine translation often mangles brand names, product names and numbers, and a wrong price or date in a subtitle is the first thing a viewer spots. Go through every proper noun and every figure in the file, one by one.

Step 3: Generate the subtitles or the dub

HeyGen is a clear example of the dubbing workflow. You upload a file or paste a YouTube link, choose the target languages (up to 10 in one job, according to its translate page), and start the translation. Before you finalize, you can open Edit and Review to proofread the translated script. HeyGen's translate page says the result comes with matched voice, lip sync and expressions, so lip sync is part of the standard output rather than an add-on. Our HeyGen review and HeyGen pricing go through the plan ladder in more detail.

Kapwing is the subtitle route. Its free plan auto-subtitles up to 10 minutes and auto-translates up to 5 minutes, so it suits one short video or a subtitle file. Its Pro plan raises the auto-translating allowance to 500 minutes a month. If you edit video in more ways than subtitles, the best AI video editors list puts Kapwing next to its rivals.

ElevenLabs is the voice-first option. Its automatic dubbing costs 2,000 credits per minute with a watermark and 3,000 without. Its Dubbing Studio costs more: 5,000 credits per minute with a watermark and 10,000 without. The pricing page lists its plans by credit pool: Starter at 30,000 credits and Creator at 121,000. The credits come from one monthly pool shared across the product, so a dub draws on the same allowance as your voice generation. Our ElevenLabs review and ElevenLabs pricing cover the rest of the plans, and cheaper ElevenLabs alternatives lists lower-cost routes to a similar voice.

Step 4: Check timing and export

Play the translated version side by side with the original. Listen hardest at the fastest-speaking passage and at any line with a number or a name. Read each subtitle at speaking pace, and if a subtitle cannot be read in the time it stays on screen, shorten the line. Fix the timing in the tool, then export.

Free exports have caps. Kapwing's free plan limits each export to 4 minutes and gives you 30 export minutes a month, with a watermark on every file. A 10-minute video therefore needs a paid plan, or you split it into three clips and export each one. Plan the export before you start, so the cap does not force a rework at the end.

What one 10-minute video costs

Take one 10-minute talking-head video, translated into one language. Here is what each route charges, using the figures on each vendor's pricing page.

• Subtitles on Kapwing. The free plan cannot cover a 10-minute video, since its translation allowance is shorter than that. Pro costs $16 a member a month billed annually. Its 500-minute auto-translating allowance covers 50 videos this length a month, if nothing else draws on it.
• Voice-over on ElevenLabs. At 2,000 credits per minute with a watermark, the video costs 20,000 credits. Without a watermark it costs 30,000 credits, which is all of the Starter pool. Dubbing Studio at 5,000 credits per minute with a watermark needs 50,000 credits, more than Starter holds, so the video needs the Creator pool of 121,000 credits.
• Lip sync on HeyGen. Creator costs $29 a month and includes 600 credits, with a maximum of 30 minutes per video, so a 10-minute video fits the limit. HeyGen's pricing page does not list credits per minute, so run a short test on your own footage before you rely on the 600 credits.
• Lip sync on Rask AI. Creator costs $39 a month billed monthly for 25 minutes, which works out to $1.56 a minute. The same 10 minutes take 40% of that month's allowance. Billed yearly, Creator is $329 for 300 minutes, or $1.10 a minute. Creator Pro, which lists lip sync on its card, costs $99 a month for 100 minutes, or $0.99 a minute.
• Translating the transcript on DeepL's free plan. The free allowance is 50,000 characters of text per user per month. Count the characters in your transcript against that figure before you start. If it fits, the translation step costs nothing on the free tier.

The cheapest route is the one that matches your output, not the lowest monthly fee. A subtitle job uses a Kapwing subscription. A dub uses a credit pool. A lip-synced clip uses a monthly plan's minutes or credits.

Gotchas that change the bill

Free tiers cap length, not only features. Kapwing's free plan caps translation and exports, and HeyGen's free plan allows three videos a month. Check the caps against your longest video before you start.

Clean files cost more. Kapwing's free exports carry a watermark on every file. ElevenLabs charges more credits for an unwatermarked dub than for a watermarked one (3,000 against 2,000 a minute), so the cleaner file uses more of your pool.

Lip sync is not on every plan. HeyGen's translate page lists lip sync as standard output. Rask's card puts lip sync on Creator Pro. ElevenLabs' pricing page does not list lip sync, so do not promise it to a client until you confirm it in the product.

Voice rights matter. A dub that uses someone else's voice needs their permission. Read the consent rules in our voice cloning guide before you clone or imitate anyone.

The source video has rights too. A translated version is still a copy of the original picture and audio. Use footage you own, or footage you have a licence to adapt, and keep that licence on file.

Names and numbers need a human pass. A translator will not know your product name, or that a price in your video is in pounds. Read the final subtitle file once, out loud, before you export.

Frequently asked questions

Can I translate a video with AI for free?

Yes, with limits. Kapwing's free plan auto-translates a short clip and exports up to 4 minutes at a time, with a watermark and 30 export minutes a month. For text, DeepL's free plan allows 50,000 characters a month. Whisper is free to download under the MIT License, but its translate option goes to English only. If you only need a voice without a translated video, the best free AI voice generators page covers the no-cost options.

How much does it cost to dub a 10-minute video?

It depends on the route. ElevenLabs' automatic dubbing uses 20,000 credits for a 10-minute video with a watermark and 30,000 without. Rask's Creator plan is $39 a month for 25 minutes billed monthly, so 10 minutes takes 40% of that month's allowance. HeyGen's Creator plan includes 600 credits and allows up to 30 minutes per video.

Does AI dubbing match the speaker's lips?

HeyGen's translate page says lip sync is part of its standard output. Rask lists lip sync on its Creator Pro card. ElevenLabs' pricing page does not list lip sync, so check the product before you promise it.

Can Whisper translate a video into Spanish?

No. Whisper's translate task goes to English only, according to its README. Transcribe the video in its source language, then translate the transcript with DeepL or another text translator.

Can I translate a YouTube video with AI?

HeyGen's translate page accepts a pasted YouTube link as an input. You still need the rights to the video and its audio before you publish a translated version of it.

Subtitles or a dub: which should I choose?

Choose subtitles when viewers will read along, such as a tutorial with on-screen steps. Choose a dub when they will listen. Before you promise lip sync to anyone, check that the plan you are buying lists it. HeyGen's translate page lists lip sync as standard output, Rask lists it on Creator Pro, and ElevenLabs' pricing page does not list it.

How we checked this

We read the pricing pages of HeyGen, ElevenLabs, Rask AI, Kapwing and DeepL, HeyGen's video translation page and the Whisper README, all in October 2026. The figures in this guide come from those pages. The per-minute and per-video costs are our arithmetic on them. We did not create a dub or a subtitle file for this guide, so the workflow steps reflect the vendors' own descriptions, not a test we ran. Where a vendor's pricing page shows promotional first-month prices that do not match its plan details, we quote the credit or plan figures that agree across the page.

Cite this: AIpresso, "How to Translate a Video With AI in 2026", October 2026.

Some offers on this page may be paid placements or contain affiliate links.