How to Dub a Training Video With ElevenLabs and What It Costs

Coding Liquids tutorial cover featuring Sagnik Bhattacharya for How to Dub a Training Video With ElevenLabs and What It Costs.
Coding Liquids tutorial cover featuring Sagnik Bhattacharya for How to Dub a Training Video With ElevenLabs and What It Costs.

To dub a training video with ElevenLabs, open Dubbing, upload the video or paste its link, choose the target languages and Speaker similarity, confirm the cost shown and download the result. The current model costs 13,500 credits per minute of video for each language on paid plans, roughly $2.20 to $2.70 a minute at monthly list prices.

The Complete Excel Course with AI Integration

From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.

By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner

Buy now

Two things trip up first-time buyers. ElevenLabs' pricing FAQ still quotes 2,000 to 3,000 credits a minute for dubbing, but those are the rates of the older v1 model; the plan tooltip for automatic dubbing on the current v2 model gives 13,500 credits a minute on paid plans. The app shows the exact charge before each dub, so trust that number. And an AI dub is a draft until someone fluent in the target language has checked it.

How much does ElevenLabs dubbing cost per minute?

ElevenLabs dubbing on the current v2 model costs 13,500 credits per minute of source video for each target language on paid plans, about $2.20 to $2.70 a minute depending on the plan.

Credits come with your monthly plan and are shared across every ElevenLabs product, so dubbing competes with any voiceovers you also make. Here is what each plan's allowance buys in v2 dubbing, worked out from the plan prices and credit allowances on ElevenLabs' pricing page (23 November 2026, monthly billing, before any promotion):

PlanMonthly priceCredits a monthv2 dubbing minutes (about)Cost per dubbed minute (about)
Free$010,000Under half a minute (22,000 credits a minute, watermarked)Not for business use
Starter$630,0002.2$2.70
Creator$22121,0009$2.45
Pro$99600,00044$2.23
Scale$2991,800,000133$2.24
Business$9906,000,000444$2.23

Three adjustments change those figures. Annual billing charges ten months for a year, so the per-minute cost drops by a sixth; ElevenLabs lists the equivalent monthly prices as $5 for Starter, $18.33 for Creator and $82.50 for Pro. Promotions come and go (a Black Friday offer on annual Creator and Pro plans runs until 1 December 2026), so budget on list prices and treat discounts as a bonus. And unused credits roll over for up to two months, to a ceiling of three times your monthly allowance, while the paid subscription stays active.

The other routes cost differently:

RouteRate on ElevenLabs' pagesWhen it fits
Automatic dubbing, v2 (default)13,500 credits a minute on paid plansMost training videos; editable in the app
Automatic dubbing, v1 (legacy)3,000 credits a minute without watermarkBudget tests; 29 languages; older model
Dubbing Studio (v1, legacy)10,000 credits a minute without watermarkClip-by-clip control and a CSV export for review; in maintenance mode
API, Dubbing v2$2.20 a minuteIf a developer automates dubbing
API, Dubbing v1$0.33 a minute with watermark, $0.50 withoutDeveloper route on the older model

The pricing page also lists watermarked v1 rates (2,000 credits a minute automatic, 5,000 in Dubbing Studio), but ElevenLabs' dubbing FAQ says watermarking isn't available on paid subscriptions, so a business plan pays the unwatermarked rates. Charges are per minute of source media for each language you select, and the app shows the total for you to confirm before the dub starts. For voiceover costs outside dubbing, see what ElevenLabs costs per minute of audio.

How do you dub a training video in ElevenLabs, step by step?

Dubbing a training video in ElevenLabs takes six steps in the app: open Dubbing, add the video, choose languages, set Speaker similarity, confirm the cost, and download when it's ready.

Before you upload

  • Lock the edit. Dub the final cut only. Every re-dub of a changed video is charged again, per minute, per language.
  • Use clean audio. One microphone per speaker, music low under speech, no talking over each other where you can avoid it.
  • Write a glossary. Product names, menu labels, acronyms and phrases that must stay in the source language (more on this below).
  • Check the limits. Automatic dubbing in the app takes files up to 180 minutes. ElevenLabs' capability page gives a 1 GB file limit and its dubbing FAQ says 2 GB, so keep files under 1 GB to be safe. The API accepts up to 3 GB per file. Accepted formats include MP4, MOV, MKV, AVI, WEBM, MP3 and WAV, with video up to 4K.

In the app

  1. Go to Dubbing in the navigation menu.
  2. Upload the video or audio file, or choose the Paste URL tab to dub from a link such as a YouTube video.
  3. In Choose languages, select one or more target languages. Each one is charged separately.
  4. Click Speaker similarity to set how closely the dubbed voice follows the original speaker, on a scale of 0 to 10. The default of 7 suits most content; higher values sound more like the presenter but can carry more of their accent and sound less natural in the new language.
  5. Click the generate button. ElevenLabs shows the cost; confirm it.
  6. When the dub is ready, open it from your list of dubs to review and edit, then download.

Getting a usable video out

ElevenLabs is explicit that the dubbed video is a high-quality preview, not a production master: it is usually output at 1080p, may not match your source resolution, and v2 audio is at most stereo. Dubbing v2 returns a single lossless audio file. For the version staff will actually watch, download the dubbed audio and replace the audio track on your original video in your editor. That takes about ten minutes per language once you've done it once, and it keeps your source quality, intro graphics and captions intact.

Dubbing works on the soundtrack. ElevenLabs doesn't offer lip sync in Dubbing, and text in the picture (slide titles, software menus, burned-in captions) stays in the original language. For training videos built from slides, plan translated slide versions alongside the dub; a fuller workflow for that is in making training videos in several languages with AI.

How does ElevenLabs handle two speakers in one training video?

ElevenLabs dubbing detects multiple speakers automatically, even when they overlap, and recreates each voice separately while keeping the original background audio under the new speech.

That matters for the common training format: a presenter introduces the topic, then a colleague demonstrates. The automatic flow's steps don't ask you to assign speakers, so check the result instead. The two failures to listen for are a short interjection ("yes, exactly") given the wrong voice, and two people talking at once producing a muddled line. Both are easier to prevent at recording time than to fix afterwards.

When you need clip-by-clip control, the legacy Dubbing Studio gives it: choose Dubbing v1 as the model, then Dubbing project. Speaker cards show each line's transcription and translation for editing; clips can be dragged to another speaker's track, split, merged or regenerated; each track can use a track clone (one voice built from all of a speaker's audio) or a clip clone. By default regenerations keep the clip's length fixed, which can speed speech up or slow it down; a Dynamic Generation adjusts the length to the text but can upset timing. Dubbing Studio handles files up to 45 minutes and 1 GB, supports fewer languages, and ElevenLabs says it's in maintenance mode with critical bug fixes only, so don't build a long-term process around it.

How should a native speaker review an AI dub?

A native speaker should review an AI dub in two passes: read the translation against the glossary, then listen to the whole dub at normal speed, logging every faulty line before fixing any.

On v2 dubs the review happens inside the app: you can edit the transcript and the translation directly, then regenerate any single segment without re-dubbing the whole video. (Editing and regenerating through the API, as opposed to the app, needs an Enterprise plan.) Before regenerating, check whether the app shows a cost for it. A reviewer's checklist, filled in for one dub:

CheckWhat to look forFound in the Spanish dub (illustrative)
RegisterFormal or informal address used consistentlySwitched between formal and informal "you" in the first two minutes; set to formal throughout
TermsGlossary terms used exactly"Phishing" translated into a literal phrase nobody uses; changed back to "phishing"
AcronymsRead as letters where they should be"MFA" pronounced as a word; written out letter by letter and regenerated
Product and menu namesLeft in the source language where the software isn't translated"Report Phish" button name translated; restored to the English label staff see on screen
Numbers and timesCorrect and in local formatFine
TimingSpeech not rushed against the next lineOne segment noticeably fast; shortened the translation and regenerated
VoiceRight voice on every lineA two-word reply from the technician in the presenter's voice; noted for the next recording

Budget review time honestly. A fluent reviewer needs to watch the whole dub at least once and the fixed segments again, so allow roughly an hour to an hour and a half per ten minutes of video for a first review. For safety, legal or compliance training, have the reviewer check the meaning of every instruction, not just the language; which business translations need a human check sets out where that line sits.

If reviewing the transcript and translation side by side in a spreadsheet is the slow part, try The Complete Excel Course with AI Integration on Coding Liquids. Dubbing Studio exports a CSV of speakers, timings, transcription and translation for that review.

A reviewer brief, filled in

Reviewers work faster with a one-page brief. The IT firm's, for its Polish reviewer (illustrative):

  • Audience: office staff at small client businesses; not technical.
  • Register: formal address throughout, the way a careful trainer would speak to staff they don't know.
  • Glossary: attached; product and button names stay as they appear on screen.
  • Priorities: meaning of every instruction first, then terms, then naturalness. Timing only where speech sounds rushed.
  • How to log: timestamp, what you heard, what it should say, severity (wrong, unclear, style).
  • Not your job: the slides; a separate translator handles on-screen text.
  • Deadline and time: two working days, up to two hours paid.

The severity column is the useful part: "wrong" lines get fixed before release, "style" lines only if time allows. Without it, a reviewer's 40 comments all look equally urgent.

A glossary prompt that saves the reviewer an hour

Give the reviewer a glossary before they start. An AI chatbot drafts one from your script in a minute:

Here is the English script of a 12-minute security-awareness training video
for office staff at small businesses. Build a glossary for dubbing it into
Spanish and Polish: list every product name, button label, acronym and
technical term; for each, say whether it should stay in English, be translated
(give the translation) or be spelled out letter by letter when spoken.
Use formal address in both languages. Flag any term where you are unsure.

A typical answer (illustrative): a 24-row table with "phishing: keep in English", "MFA: spell out, M-F-A", "Report Phish button: keep the English label", "password manager: translate", and a note that "ticket portal" could be translated two ways. What to fix: the chatbot proposed a translation for the firm's own product name, which must stay as it is, and it marked "VPN" for translation when both reviewers said staff say "VPN". Treat the chatbot's glossary as the reviewer's starting list, not a decision.

Dubbing a 12-minute training video into two languages: the numbers

Here is a whole job for an illustrative IT support firm. The firm made a 12-minute security-awareness video for its clients' office staff: the owner presents, and a technician demonstrates how to report a suspicious email. Two client businesses asked for Spanish and Polish versions for staff who are more comfortable in those languages.

ItemSumResult
Dubbed minutes12 minutes × 2 languages24 minutes
Credits for the first pass24 × 13,500324,000 credits
Allowance for regenerating fixed segmentsAbout 10% extra, as a marginAbout 32,000 credits
Plan that covers it in one monthCreator's 121,000 credits fall short; Pro has 600,000Pro, $99 for the month
Same job via the API24 × $2.20$52.80, if a developer runs it
Review timeAbout 1.5 hours per language for 12 minutes3 hours of reviewers' time
Editing timeReplace audio tracks, export two versionsAbout 30 minutes

The firm took one month of Pro for $99, used about 356,000 credits on the job, and kept the rest for the next video while the subscription ran. Had it dubbed on Creator, it would have needed the extra credits for 15 more minutes of dubbing on top of the plan. Switching to annual Pro billing made sense only once the firm knew it would dub every month.

A cheaper test is worth running first. The firm dubbed the first two minutes on Starter ($6, about 2.2 minutes of v2 dubbing) to hear whether the owner's cloned voice sounded right in Polish at Speaker similarity 7. It didn't quite: the accent came through strongly. At 5 the delivery sounded more natural and still recognisably the owner, so the full dub used 5. Two minutes of testing saved a full re-dub at roughly $53 of credits.

If you want help fitting dubbing and review into how your business produces training, I offer one-to-one AI consultations for small businesses.

Is dubbing or a translated voiceover cheaper for a training video?

A translated voiceover made with ElevenLabs text to speech usually costs far less than a dub: text to speech uses about 1 credit per character, v2 dubbing 13,500 credits a minute.

The trade is delivery against cost. A dub keeps the presenter's own voice, pacing and the second speaker's lines in place. A voiceover reads a translated script in one chosen voice, and you lay it over the slides yourself. Here are the options for the IT firm's 12-minute video, per language (credits from ElevenLabs' pricing page; the script length is an estimate at a typical narration pace of about 150 words a minute, roughly 11,000 characters):

OptionElevenLabs credits per languageKeeps the presenter's voiceExtra work
Dub with v212 × 13,500 = 162,000YesReview and audio swap
Dub with legacy v112 × 3,000 = 36,000Yes, older modelReview, audio swap; 29 languages only
Transcribe, translate, then text to speechAbout 4,000 for the transcript (330 credits a minute) plus about 11,000 for the voiceoverOnly with a cloned voiceTranslate the script, re-time the voiceover to the slides
Subtitles onlyAbout 4,000 for the transcriptOriginal audio staysTranslate and add a caption file
Re-record with a bilingual presenterNoneA different person's voiceA recording session per language

On the Pro plan's rates, the v2 dub costs about $27 of credits per language for this video and the voiceover route about $2.50, an elevenfold difference. That changes the decision for slide-only training with one narrator: translate the script (with the reviewer checking it before any audio exists), generate the voiceover, and drop it onto the slides. Choose the dub when the presenter is on camera, when two people talk to each other, or when the video will be watched by clients who know the presenter's voice. Subtitles suit staff who follow the source language well enough and videos that change often, since editing a caption file is free. Whichever route you pick, the review step stays: a translated script read in a confident AI voice is just as wrong as a dub if the translation is.

Which ElevenLabs plan suits a business dubbing training videos?

The right ElevenLabs plan for dubbing is the smallest whose monthly credits cover your dubbed minutes, counting each language separately; Starter is the minimum because it adds the commercial licence.

Dubbed minutes a month (all languages)PlanWhy
Testing only, under 2StarterCommercial licence; enough for a short sample
Up to about 9CreatorOne short video in one language a month
Up to about 44ProA 10 to 20-minute video in two or three languages
Up to about 133ScaleA regular programme; three workspace seats for reviewers
Up to about 444BusinessA large library; ten seats

Seats matter if reviewers need their own logins: Scale includes three workspace seats and Business ten. If you dub in bursts, rollover lets a quiet month's credits carry forward for up to two months. Plan differences beyond dubbing, such as Professional Voice Cloning arriving on Creator, are covered in ElevenLabs Starter vs Creator for a small business.

What goes wrong when dubbing training videos with AI?

Most AI dubbing problems in training videos come from dubbing too early, forgetting the per-language charge, untranslated on-screen text, mispronounced acronyms and treating the preview video as the finished file.

The acronym that came back as a word

In the IT firm's first Polish dub, "MFA" came out as a single invented word, three times in the section that matters most. Nobody noticed until a client's office manager asked what it meant. The fix took five minutes: in the translation, write the acronym as separate letters (or expand it to the full term the first time), regenerate those segments, and add every acronym to the glossary with an instruction. It's the commonest small fault in technical training dubs, and the cheapest to prevent.

  • Dubbing a draft. A late edit to a 12-minute video means another full dub in every language. Dub after sign-off.
  • Counting minutes once. Three languages triple the bill. Multiply source minutes by languages before you choose a plan.
  • Slides left untranslated. The voice speaks Spanish while the slide says "Step 3: Verify the sender". Translate slides or add translated captions.
  • Shipping the preview. The dubbed video is a 1080p preview with at most stereo audio. Swap the dubbed audio into your original file instead.
  • No presenter consent. Dubbing recreates a real person's voice in a language they don't speak. Get written agreement, especially from staff who might later leave.
  • Building on Dubbing Studio. It's in maintenance mode. Use it for one-off control, not as the backbone of a long-term workflow.
  • Using the Free plan for real videos. Free dubs are watermarked and the plan has no commercial licence.

For a comparison with an avatar-video tool's translation feature, see translating a video with HeyGen and counting the credits, and for corporate video more broadly, AI video translation for corporate video.

ElevenLabs dubbing: costs, limits and quality

Is ElevenLabs dubbing free?

Only for a short test. Dubbing is available on the Free plan, but free dubs are watermarked automatically and the Free plan has no commercial licence. The current dubbing model costs 22,000 credits per minute on Free, and the plan includes 10,000 credits a month, so you can dub less than half a minute. For a business training video, Starter is the first plan with a commercial licence.

How many languages can ElevenLabs dub a video into?

More than 90 with the current Dubbing v2 model: ElevenLabs' docs say 90+ and its API pricing page lists 92. The older v1 model, which the legacy Dubbing Studio uses, supports 29. Each target language is charged separately for every minute of source video, so dubbing a 10-minute video into three languages is billed as 30 minutes.

Does ElevenLabs dubbing keep the presenter's own voice?

Yes, it aims to. Dubbing recreates each speaker's voice and delivery in the new language, and the Speaker similarity setting, on a 0 to 10 scale with 7 as the default, controls how closely. Higher values sound more like the original person but can carry more of their accent and sound less natural. Get the presenter's written agreement before you recreate their voice.

Does ElevenLabs dubbing include lip sync?

No. ElevenLabs' dubbing FAQ says lip sync isn't part of Dubbing; it's offered in other ElevenLabs tools through third-party models. For training videos this rarely matters, because most screens show slides or software rather than a close-up face. Where the presenter is on camera for long stretches, plan to cut away to slides during the dubbed sections.

Do unused ElevenLabs credits roll over?

Yes, for up to two months on paid plans, with a ceiling of three times your monthly allowance, as long as the subscription stays active and you don't downgrade or cancel. That helps a business that dubs in bursts: credits saved in quiet months can cover a bigger video later. Rollover doesn't apply to the Free plan.

Further reads

Sources: ElevenLabs pricing page (plan prices, credits, rollover and the automatic dubbing credit tooltip), API pricing page, Dubbing capability docs, Dubbing product docs and FAQ, Dubbing Studio docs and the help centre article How much does Dubbing cost, all read on 23 November 2026.

Rather not work this out alone?

An AI implementation consultation is a one-to-one call with me, not an agency or a sales team. You leave knowing what AI you can use in your business and where to start.

Book a 1:1 call with me