Key Takeaways
- You translate a video in three moves: transcribe the audio, translate the transcript against a glossary, then deliver it as subtitles, a voiceover, or a dubbed track.
- Translation is not localization. Localization also adapts on-screen text, graphics, dates, currency, idioms, and examples so the video feels made for that market.
- Method decides cost and feel. Subtitles are the cheapest and fastest option and work with the sound off. Voiceover suits explainers. Dubbing with lip-sync feels the most native but costs the most, and viewer preference for it splits sharply by country.
- Don't ship 12 separate videos for 12 languages. Deliver one video with multiple audio and subtitle tracks and let the viewer pick their language in the player.
- The free, built-in option is often the weakest one. A YouTube creator tested a dubbed audio track against a dedicated localized channel with the same video, and the audio-track version pulled 32 views in 90 days while the dedicated channel outperformed it by over 100 times.
- Subtitles should be the default starting point for every video before you spend on dubbing, because a meaningful share of video is watched with the sound off.
How do you translate a video so it actually works for someone in another country? You transcribe the audio, translate the transcript against a glossary instead of word-for-word, then choose a delivery method: subtitles, voiceover, or dubbing.
That's the mechanical answer. The part most articles skip is what happens after the words are translated, and how you serve five languages from one video without maintaining five separate uploads.
The same workflow holds whether you're translating a video to English from another language or taking an English original into ten markets at once.
This is for marketers, creators, course builders, and founders who have a video sitting in one language and want it working in two or three more, this week, without guessing at which method to use or what it costs.
Video translation swaps the words. Video localization adapts the whole experience: on-screen text, currency, idioms, even color and gesture choices that read differently outside their home market. Confusing the two is the single most common reason a "localized" video still feels foreign to the audience it was made for.
This article will walk you through the full workflow: transcribe, translate, choose a method, localize beyond the words, produce the asset, deliver it in one player, then make it discoverable and measurable per market.
By the end you'll know which named tools handle which step, what each step actually costs, and why the last two steps, delivery and discoverability, are where most localization projects quietly lose their return.
Translation vs. Localization: What's the Difference?
Translation converts spoken or written words from one language into another. Localization does that and then adapts everything else: on-screen text, graphics, colors, dates, currency, units, idioms, humor, and cultural examples, so the video reads as if it were built for that market instead of retrofitted for it.
A video can be translated perfectly and still fail on localization. A price shown in U.S. dollars, a reference to a U.S. holiday, or an idiom that has no equivalent in the target language will all read as foreign, even with flawless subtitles under them. The words are right. The frame around them isn't.
Translation makes a video understandable. Localization makes it feel like it was made for you.
Once that distinction is clear, the next decision is mechanical: which delivery method fits the content and the budget.
Subtitles vs Voiceover vs Dubbing: How to Choose
Pick the delivery method by content type, budget, and how much the original voice matters to the message. Not every video needs the same treatment, and defaulting to the most expensive option is usually the wrong call.
Viewer preference for dubbing versus subtitles splits sharply by country, and the split has held for years.
A 2022 Morning Consult survey of adults across 15 countries found that respondents in Russia, Germany, Italy, Spain, and France largely preferred dubbing, while roughly 7 in 10 adults in China and South Korea said they preferred subtitles. The United States leaned slightly toward subtitles over dubbing.
That regional pattern is the closest thing to an industry constant in this space, and it should shape which markets get dubbing budget first.
Subtitle everything, dub selectively, and pay for lip-sync only where the face and the voice are the product.
As a rule of thumb, subtitles cost a fraction of a dollar per minute with AI tools, voiceover sits in the low tens of dollars per minute, and lip-synced dubbing is the only method where a per-minute budget conversation is even necessary.
How to Translate and Localize a Video, Step-by-Step
Eight steps, start to finish. Each one names the tools that actually do the job.
Step 1: Transcribe the Source Video First
Everything downstream depends on an accurate transcript. Translate a bad transcript into five languages and you scale the same five mistakes.
Auto-generate a transcript first, then clean it by hand: fix names, brand terms, and technical vocabulary before anyone starts translating. Build a glossary or termbase up front, covering product names, people's names, and units, so every language version stays consistent instead of drifting language by language.
Platforms with built-in AI transcription skip the manual upload-to-transcript step entirely. Gumlet transcribes audio in 30 languages as part of its subtitle and caption generator, which covers the languages most video-heavy SaaS and marketing teams actually need. Descript and Maestra offer comparable transcription tooling if you're evaluating alternatives.
Fix the transcript once, or fix the same mistake in every language later.
Step 2: Translate the Transcript, Not Word-for-Word
Literal translation breaks on idioms, tone, and timing. The goal is translating for meaning, then adapting, not swapping one word for its dictionary equivalent.
Translate against the glossary you built in step one. Adapt idioms and examples rather than carrying them over untouched.
Watch line length: According to Kwintessential’s 2025 Translation Text Expansion report, several major European languages expand by 15% to 30% when translated from English, with German expanding up to 35%, so a subtitle line that fits the original timing may overrun it in the translated version.
AI translation is fast, but it isn't quality assurance. Enforce the glossary and have a native speaker review names, numbers, and technical terms before anything ships. Gumlet translates subtitle and caption text into 90-plus languages, which covers the text layer well. It doesn't touch voice.
Step 3: Choose Your Method Per Video
Decide the delivery method by content type and target market before spending any production budget, using the comparison table above as the starting filter.
Default to subtitles for reach and for viewers watching with the sound off. Add voiceover for explainer and e-learning content where narration carries the instructional weight. Reserve dubbing and lip-sync for face-heavy content or markets that skew toward dubbing, per the regional data in the previous section.
Dubbing costs an order of magnitude more than subtitles even with AI pricing. Spend that budget where the voice itself is part of the value, not on every video by default.
Step 4: Localize Beyond the Words
This is the step that separates a translated video from a localized one: adapting everything the words alone don't cover.
Swap on-screen text and lower-thirds for the target language. Adjust dates, currency, units, and address or phone formats to local convention. Replace culturally specific imagery, examples, or humor that won't land outside its home market. Recheck colors and gestures, since meanings shift by region in ways that rarely show up until a local reviewer flags them.
As an illustrative case rather than a documented one: a marketing video that tests well in one German-speaking market can flop in a neighboring one that shares the language, purely on cultural fit rather than translation quality.
That gap is exactly what step four exists to catch, and it's invisible if you only ever check the subtitle file.
A short pre-launch checklist keeps this step from getting skipped under deadline pressure:
- On-screen text and lower-thirds translated
- Dates and currency converted to local convention
- Culturally specific imagery or humor reviewed by a native speaker
- Colors and gestures checked against local meaning
- A final pass by someone who lives in the target market rather than someone who just speaks the language.
If your video still shows a U.S. dollar sign in the German cut, you translated it, but you didn't localize it.
Step 5: Produce the Localized Asset
This is where the actual subtitle files, voiceover tracks, or dubbed audio get generated, and where the AI dubbing tools live.
Export subtitles as SRT or VTT. For voice, run the source through an AI dubbing tool, then route the output to a human reviewer before it ships. Keep the original video master untouched so you can re-render if a language needs a fix later.
The current tool landscape, by use case:
- HeyGen supports 175-plus languages with lip-sync and avatar options, and is the strongest fit for creators who want lip-synced translation without a studio.
- ElevenLabs released Dubbing v2 in 2026, still in alpha, expanding its dubbing coverage to 90-plus languages while preserving the original speaker's tone, pacing, and emotional delivery, a sharp jump in reach from its earlier dubbing tools.
- Rask AI covers 130-plus languages and is built for batch and scale work across a video library, with voice cloning available in 32 of those languages.
- Synthesia handles avatar-based video generation.
- YouTube's built-in auto-dubbing in Studio is the fastest way to test a language before committing a budget anywhere else, and as of February 2026 it's available to all eligible creators across 27 languages.
Gumlet is not a dubbing tool. It doesn't clone voices and it doesn't do lip-sync. Its job starts at the next step, delivery, and that's a deliberate scope, not a gap.
For a broader roundup of options beyond the tools named here, see our guide to the best AI video translation tools.
Step 6: Deliver Every Language in One Player
This is the step most guides skip entirely. Once translated tracks exist, don't publish a separate video per language. Host one video with switchable audio and subtitle tracks instead.
Upload a single master file. Attach each translated subtitle track and each dubbed audio track to that one asset. Expose a language picker inside the player itself. Set a sensible default track per region so the right language plays automatically for most viewers.
Publishing a separate video per language multiplies maintenance work, splits your analytics across duplicate assets, and fragments your SEO across near-identical pages competing with each other.
One asset carrying many tracks avoids all three problems at once. Gumlet supports multiple audio tracks and multiple subtitle tracks on a single video asset, delivered through adaptive streaming, which lets a viewer switch language inside the player without loading a different URL. This is close to the natural home turf for a video hosting platform, versus raw file hosting that has no concept of a track at all.
Ship one video with 10 tracks, not 10 videos. Your future self, maintaining the library 18 months from now, will thank you.
Step 7: Make Localized Videos Discoverable
A translated video nobody can find, in their own language, is a wasted production budget. Every market needs a path to the content, not just a track buried inside it.
For web pages, build language-specific pages with translated titles and descriptions, and implement hreflang correctly so search engines serve the right language version to the right visitor.
A minimal hreflang setup for a three-language video page looks like this in the page's <head>:
Each version must reference all the others, including itself, and the x-default line tells search engines which page to show a visitor whose language doesn't match any listed version. Missing the self-reference or the reciprocal links is the most common way hreflang setups silently fail.
For YouTube specifically, use translated video metadata and multi-language audio tracks together. For markets that justify the investment, a dedicated localized channel or page will outperform a buried audio track by a wide margin.
Here's the clearest evidence for that last point. Lucas Conde, a Kapwing-affiliated YouTube creator with 162,000 subscribers, ran a controlled test in early 2026: he dubbed the same viral video into Spanish and published it two ways on the same day.
One version went up as a dubbed audio track on his existing main channel, using YouTube's built-in multi-language audio feature. The other went to a brand-new, dedicated Spanish-language channel.
After 90 days, the audio-track version on the main channel had pulled 32 views. The dedicated channel had 3,897, over 100 times as many.
The gap wasn't translation quality. It was that a dedicated channel points every discovery signal, language, audience, metadata, in the same direction, while a buried audio track leaves YouTube's algorithm trying to reconcile two audiences on one channel.
Translated captions also feed both search engines and AI answer engines directly, so pairing Gumlet's translated subtitle text with properly localized per-language pages and correct hreflang tagging compounds the discoverability effect rather than relying on the video platform alone.
Step 8: Measure Per Market and Iterate
Localization return on investment is market-specific. An aggregate view of watch time or completion rate hides which languages are actually earning their production cost and which aren't.
Track watch time, completion rate, and conversion broken out by language and region rather than as one blended number. Compare method performance directly: did dubbing actually outperform subtitles in a given market, or did you spend the dubbing budget on assumption rather than evidence?
Double down where retention is strong. Cut the languages where it isn't, even if the aggregate number still looks fine.
A video can win in one market and lose in another that technically shares the same language. Segment the analytics before drawing a conclusion, not after.
A market-by-market lens like this is what separates a localization program from a translation habit. The dubbed German audience that completes at a noticeably higher rate than the subtitled version tells you exactly where the next dubbing budget should go; the market that shows no lift tells you just as clearly where to stop spending.
Localization you don't measure per market is just translation you're hoping worked.
Common Mistakes That Ruin Video Localization
Most localization failures trace back to a short list of repeatable errors:
- Translating literally, word-for-word, instead of translating for meaning and adapting from there.
- Skipping the glossary, which lets names, product terms, and technical vocabulary drift across languages.
- Translating the spoken words while leaving on-screen text, currency, and cultural examples untouched.
- Dubbing everything by default, including content where lip-sync adds cost without adding value.
- Publishing a separate video per language instead of one asset with multiple tracks, multiplying maintenance, and fragmenting analytics.
- Skipping native-speaker review on AI-generated output before it goes live.
- Treating subtitles as optional instead of as the baseline every other method builds on top of.
Most "our localized video flopped" stories are really "we translated the audio and forgot everything else on screen."
How to Start This
Pick your single highest-value video and your top one or two target markets. Ship subtitles before you touch dubbing.
The sequence:
- Transcribe and clean the source video
- Translate into your top one or two languages against a glossary
- Publish as subtitle tracks first
- Measure retention by market
- Invest in voiceover or dubbing only for the markets that proved to have high viewership
- Deliver everything through one multi-track player rather than separate uploads
Teams testing this workflow can auto-transcribe, translate captions, and serve multiple subtitles and audio tracks from a single video through a free plan before paying for any dubbing tool at all.
Closing Thoughts
Translating a video is the easy half of this problem. Transcribe it, translate the transcript against a glossary, pick a delivery method that fits the content and the market.
The harder half, and the part every competing guide skips, is localizing beyond the words and then delivering every language from one asset instead of a pile of duplicate uploads that fragment your analytics and your SEO at the same time.
The principle holds regardless of which tools you use: transcribe first, translate for meaning, localize beyond the words, deliver in one player, then measure per market instead of in aggregate. Skip any one of those five and the localized version underperforms the English original, not because the translation was wrong, but because the workflow around it was incomplete.
If you're ready to test this on your highest-value video, Gumlet's free plan covers transcription, subtitle translation, and multi-track delivery, enough to validate the first two markets before you spend anything on dubbing.
Frequently Asked Questions
1. How do you translate a video?
Transcribe the source audio into an accurate text file, translate that transcript against a glossary rather than word for word, then deliver the result as subtitles, a voiceover, or a dubbed audio track. Subtitles are the fastest and cheapest starting point for most videos. The transcript quality determines everything downstream, so clean it before translating rather than after.
2. What's the difference between video translation and localization?
Translation converts the spoken or written words into another language. Localization does that and also adapts on-screen text, graphics, currency, dates, idioms, and cultural examples so the video feels built for that market rather than translated into it. A video can be perfectly translated and still fail on localization if the visual and cultural details underneath the words go untouched.
3. Should I use subtitles or dubbing?
Default to subtitles for most videos, since they're the cheapest option and work for viewers watching with the sound off. Reserve dubbing for face-heavy content, branded marketing videos, or markets where survey data shows a strong dubbing preference, such as Germany, France, Italy, and Spain. Spend the dubbing budget only where the voice itself carries part of the message.
4. How much does it cost to translate a video?
Subtitles are the cheapest method by a wide margin. AI dubbing typically runs $2 to $20 per finished minute depending on the platform and volume, compared with $100 to $500 per finished minute for traditional studio dubbing with human voice actors. Voiceover without lip-sync falls between the two. Exact cost depends heavily on language pair and whether lip-sync is included.
5. What are the best AI video translation tools?
HeyGen supports 175-plus languages with lip-sync and avatar generation. ElevenLabs' Dubbing v2, released in 2026 and still in alpha, covers 90-plus languages while preserving the original speaker's tone and pacing.
Rask AI covers 130-plus languages and is built for batch localization across a video library. YouTube's built-in auto-dubbing is free and available to eligible creators in 27 languages as of February 2026, making it the fastest way to test a language before committing a budget elsewhere.
6. How many languages does YouTube's auto-dubbing support?
27 languages as of mid-2026, when YouTube opened auto-dubbing to all creators. Expressive Speech, which carries the creator's original tone and energy into the dub, launched in eight of those languages. It's free, which makes it the cheapest possible way to test whether a market responds to your content before spending on a dedicated dubbing tool.
7. How do I show a video in multiple languages without uploading it multiple times?
Upload one master video and attach multiple audio tracks and subtitle tracks to that single asset instead of publishing a separate file per language. A player that supports multi-audio and multi-subtitle delivery, such as Gumlet's, lets viewers pick their language inside the player.
This avoids the maintenance burden, analytics fragmentation, and SEO cannibalization that come from publishing duplicate videos per language.
8. Does translating a video help SEO?
Yes, when the translation extends past the video file itself. Translated captions feed both search engines and AI answer engines directly, and pairing them with language-specific pages using correct hreflang tags helps each market's search results surface the right version.
A translated audio track with no supporting metadata or page structure captures far less of this benefit than a properly localized page does.
9. Can I translate a video for free?
Yes, within limits. Gumlet's free tier includes unlimited AI-generated subtitles and multi-language audio and subtitle delivery, and YouTube's built-in auto-dubbing is free for eligible creators.
Free tools cover transcription, translation, and basic delivery well, but AI-generated output still needs a native-speaker review pass before it goes live, since quality assurance is the part free tiers don't include.
10. What is hreflang and do I need it for a translated video?
Hreflang is an HTML tag that tells search engines which language and region a page is meant for, so a German searcher lands on the German version of your video page instead of the English one. It matters most when you've built separate landing pages per language rather than relying on in-player track switching alone. Skipping it is one of the most common reasons a translated page never surfaces in the right market's search results.
