Research-backed calculator guide

Video Translation vs Dubbing Cost: Choose Subtitles, Voiceover or Lip-Sync

Compare video translation, subtitles, voiceover, AI dubbing, and lip-sync costs with a transparent per-minute method, worked example, and review assumptions.

What “translation” and “dubbing” include

Video translation can mean only a translated transcript, a subtitle file, or the complete preparation of a localized version. Those are different deliverables. A translated script is text. Subtitles also require timecodes, line breaks, reading-speed decisions, file validation, and often a native-language review. A dubbed video needs a target-language audio track, and it may need music and effects separated from the original dialogue before the new track is mixed.

Voiceover and dubbing are also not identical. In voiceover, the original speech may remain quietly audible underneath the translated voice. In replacement dubbing, the original speech is removed or heavily reduced. Lip-sync is the strictest version of replacement dubbing: the translated performance must fit visible mouth movements, shot timing, pauses, and sometimes facial expression. That extra constraint creates adaptation and revision work even when the synthetic voice itself is inexpensive.

The useful unit is the finished localized minute, not simply “one video.” A six-minute video in four languages is 24 localized minutes before any revision. If a source cut changes twice, you should model whether the change touches all 24 minutes or only a few segments. Re-recording or re-rendering the whole version is a very different budget from updating a subtitle line and one short audio segment.

How to compare subtitles, voiceover, and lip-sync

Use four questions before looking at rates.

1. What must the viewer do? Subtitles ask the viewer to read while watching. Voiceover lowers that reading burden. Lip-sync aims for the most immersive result.
2. What is on screen? If a presenter’s face is visible for most of the video, mismatched timing is more noticeable. If the video is a screencast or slide presentation, subtitles or voiceover may be enough.
3. What is the cost of an error? A casual internal update can tolerate a targeted correction pass. A regulated, medical, legal, or flagship brand video needs qualified review and stronger approval controls.
4. How often will the source change? Frequent updates favor a workflow that can revise text and regenerate short segments. A studio performance with new casting or booking requirements may make each revision expensive.

This decision does not make a human quote unnecessary. It tells you what to ask a vendor to itemize: translation or adaptation, voice talent, AI minutes or characters, lip-sync, sound mix, review, rights, project management, rush work, and revisions.

The meters behind an AI estimate

Official API meters help you isolate the machine portion of a budget.

• ElevenLabs documents dubbing as a per-minute-of-source-media charge for each language. It also says the API creation charge prepays the first language and that additional languages are charged separately when queued. ^[1]
• On the 2026-11-02 check, ElevenLabs listed Dubbing v1 at $0.33 per source minute with a watermark and $0.50 without one, and Dubbing v2 at $2.20 per minute. The same page listed text-to-speech at $0.04 or $0.08 per 1,000 characters and Scribe v2 speech-to-text at $0.22 per hour. ^[2]
• Google Cloud Speech-to-Text says audio is billed in one-second increments and that multiple channels are billed separately. Its standard Speech-to-Text V2 table showed $0.016 per minute for the first 500,000 minutes in a month, while its dynamic-batch standard rate showed $0.003 per minute. ^[3]
• Google Cloud Text-to-Speech bills input characters, including spaces and newline characters. Its pricing page showed, among other tiers, $4 per million characters for Standard or WaveNet voices after the listed free allowance and $16 per million for Neural2. ^[4]

These figures are not interchangeable. A dubbing product may bundle transcription, translation, speaker handling, synthesis, timing, and a video export. A build-your-own pipeline may pay separate STT, translation-model, TTS, storage, rendering, and review costs. Do not add a standalone STT price to an all-in-one dubbing rate unless you know the product already excludes transcription.

Worked cost example

Assume a six-minute English training video, two target languages, one first release, and a human review pass. The following is an estimate, not a vendor quote. The machine rates are dated official list prices above. Review and editing are explicit planning assumptions so that another team can replace them with its own hourly rates.

The table combines official machine-meter assumptions with deliberately labeled planning assumptions. It does not claim that a provider will sell subtitles for $2 per minute or that every lip-sync workflow costs $2.20 per minute. The $2.20 figure is an official Dubbing v2 meter; the review, mix, and other rows are the parts a buyer must confirm.

The calculation is straightforward. Twelve localized minutes require $0.02 of transcription at the stated Scribe assumption, plus $0.58 of TTS characters, plus the method-specific row. Subtitles get the lightest review and no new audio mix. Voiceover gets more review and a mix. Lip-sync gets the largest review and export allowance because timing and visual synchronization need a full listening-and-watching pass.

Now model two source updates. If each update changes the full six-minute video, multiply the first-release totals by three: $148.80 for subtitles, $331.80 for voiceover, and $378.00 for lip-sync. If only 30 seconds changes in each language, do not multiply blindly. Reprice 1 localized minute of changed content, then add any minimum job charge, reviewer setup, re-export, or studio booking fee. This is why “price per video” can conceal the main cost driver.

For a reusable component comparison, put the rates and unit conversions into the Model pricing comparison. If your workflow also calls a language model to adapt scripts or check terminology, use the AI token cost calculator and keep those token costs separate from audio and video production. A broader LLM cost calculator can help model the surrounding automation, but it does not replace a vendor quote for talent, rights, or final mastering.

When subtitles are the right cost choice

Choose translated subtitles when the audience can comfortably read, the original performance matters, or the video contains screen content that would be difficult to dub. Subtitles generally avoid voice casting, recording, speech synthesis, and a new dialogue mix. They can also be revised at the text layer when the source changes.

Do not treat subtitles as an automatic zero-review option. Names, numbers, product terms, line length, reading speed, and timing still need checking. A transcript error repeated in five languages is a cheap machine output that becomes expensive rework later. Budget a native-language review, especially for customer-facing, technical, or compliance content.

When voiceover is worth the premium

Voiceover is often the practical middle path for training, interviews, narration, and explainers. It reduces reading effort while avoiding the strictest mouth-shape constraint. A synthetic voice may reduce recording time, but the project still needs pronunciation guidance, timing checks, a clean mix, and a decision about whether the original speaker remains audible.

Voiceover can also be a good compromise for a mixed library. Use subtitles for short social clips and interface-heavy tutorials, then use voiceover for long training modules where reading every sentence would slow the learner. Model each content type separately instead of assigning one method to every video in a catalog.

When lip-sync earns its cost

Lip-sync is most defensible when faces and performance carry the message: acted scenes, a flagship brand film, a presenter addressing the camera, or a market where immersion is a requirement. It is usually harder to revise because the adapted line must fit both the translation and the visible action. It can also require separated dialogue, music, and effects tracks, plus more detailed review.

Lip-sync is not automatically the best choice because it looks advanced. If the viewer’s attention is on a software interface, a translated subtitle track may communicate more clearly and cost less. Reserve lip-sync for the videos where the visual synchronization changes the viewing experience enough to justify the added production and QA work.

A practical budget checklist

Before accepting a quote, ask whether the rate is per source minute or finished minute, whether each target language is charged separately, and whether partial minutes are rounded. Confirm whether transcription, translation, voice generation, speaker separation, lip-sync, captions, mixing, export, and review are included. Ask how updates are priced and whether a failed generation is refunded or retained as credit. ElevenLabs, for example, documents refund handling for some failed project or language-generation cases, but product rules differ and should be checked before purchase. ^[1]

Also record assumptions. State source runtime, number of speakers, target languages, language variants, expected character count, review hours, rights and consent requirements, delivery format, and revision count. Keep API rates and labor assumptions in separate rows. That makes a later invoice reconciliation possible and prevents a low machine-meter number from being mistaken for a finished localized-video budget.

FAQ

Is video translation cheaper than dubbing?

Usually, translated subtitles are cheaper because they add text timing and review without recording or synthesizing a new audio track. Voiceover and dubbing add audio production. Lip-sync adds the strictest timing and visual checks. The exact difference depends on language count, speakers, source quality, review depth, and revisions, so use a component estimate rather than a universal multiplier.

What is the cheapest way to localize video?

For a readable audience and a short-lived or operational video, translated subtitles are often the lowest-cost complete treatment. A transcript-only translation is cheaper still, but it is not a finished viewer-ready asset. Include timing, line fitting, QA, captions, and export before comparing it with voiceover or dubbing.

Is AI dubbing cheaper than human dubbing?

AI dubbing can reduce machine-production and turnaround costs, but the comparison is fair only when review, adaptation, mixing, usage rights, and revisions are included on both sides. Official AI meters are estimates of product usage, not proof of a publish-ready result. High-stakes or performance-led content may still need human translation, direction, talent, or approval.

Should I use voiceover or lip-sync?

Use voiceover when the viewer needs to listen but the speaker’s mouth is not the main visual focus, such as training narration or an interview. Choose lip-sync when visible speech and performance are central enough that timing mismatches would damage comprehension or brand quality. If you are unsure, price both with the same source minutes, languages, and review standard.

How should I budget revisions?

Separate full-master changes from segment changes. A new six-minute cut in two languages may repeat the entire localized workload. A changed 30-second segment may need only a small translation, regeneration, timing check, and export, although minimum charges can apply. Track revision count and changed minutes as inputs rather than hiding them in a contingency percentage.