ABSTRACT
01.The four ways to translate a video
Every method answers the same question differently: how much of the original performance survives the language change? The words are the floor. Above them sit your voice, your face, and the scene around you: music, effects, ambience, the laugh that lands mid-sentence.
| Method | The words | Your voice | Your face | The scene | Typical price |
|---|---|---|---|---|---|
| Subtitles | Read, not heard | Kept (original audio) | Kept | Kept | Free on every Familiar plan · professional translation about $10/min |
| Voice-over (the classic dub) | Spoken by a stranger | Replaced | Mouths the original language | Depends on the studio mix | $39 to $94 human · from $128 premium studio |
| Voice-only dubbing | Spoken | Yours | Still mouths the original | Often degraded (measured in Sec 02) | $3 (ElevenLabs checkout quote, Aug 5, 2026), audio only |
| Voice + face translation | Spoken | Yours | Re-rendered to the new language | Music, effects, ambience kept | $2.50 complete (Familiar) |
Subtitles keep everything but ask the viewer to read while watching. The classic dub trades your voice away: YouTube auto-dubbing replaces your voice with a stock stranger voice, and human voice-over does the same with a paid actor. Voice-only dubbing, often searched as "voice cloning," keeps the voice but leaves the mouth speaking the old language. Voice + face translation translates the performance itself; how that works under the hood is covered in AI Video Dubbing, Explained.
02.What quality means and how it is measured
The honest test is paired and same-clip: send the identical source video to two systems, compare the finished outputs on the same dimensions. Four dimensions matter: meaning (did the words survive?), voice identity (does it still sound like you?), scene audio (did the music and reactions survive?), and timing (do lines land where the original landed?).
Two paired studies against ElevenLabs, each reported separately, live at /benchmark. On the stable Dubbing v1 API study (418 paired outputs, 11 languages), Familiar made 48.8% fewer review-flagged spoken-output mistakes (132 vs 258) and measured 27.8% higher speaker resemblance, with all 11 languages favoring Familiar. On the newest-product study (Website Dubbing V2 Alpha, 111 paired outputs), ElevenLabs made 121% more important spoken-meaning mistakes and 270% more background-sound error.
03.What translation quality is achievable
Translation text itself can be scored against professionals. On WMT24++ passages, Familiar's faithful translations reached 96% of a first-pass professional translator's COMET score (95% CI: 95.2 to 96.9%), under a reference design that favors the professional baseline. Per language, Hindi scored above the professional first pass, French tied it, and Indonesian reached 99.6%. A production-pipeline re-run is in progress; we expect the production system to match or beat first-pass professionals.
The model is only half of translation quality; control is the other half. A Do Not Translate list keeps names and catchphrases exactly as you say them, jokes get rewritten to land rather than translated word for word, and videos pause at a review screen where you can edit any line before render.
04.Cost per minute
| Option | Per minute | What it covers |
|---|---|---|
| Familiar complete dub | $2.50 | Translation, your voice, scene audio, the face re-rendered |
| Familiar live dubbing | 42¢ | Streams translated in real time, per language channel |
| ElevenLabs | $3 (checkout quote, Aug 5, 2026) | Audio only |
| Professional subtitle translation | About $10 (median, 11 study languages) | Text only: translation, timing, QC |
| Human dubbing packages | $39 to $94 | Voice actors, audio only |
| Premium studio dubbing | From $128 | Studio cast and mix |
A 60-minute video in one language is $150 complete, and subtitles in the dubbed language are free on every Familiar plan, downloadable as .srt or .vtt. The free tier covers 4 minutes of video a month in 1 language with no card; paid plans start at $35/mo and open all 24 target languages. Plan mechanics are on /pricing and in the FAQ; the full rate survey with sources is in How Much Does Dubbing Cost in 2026?
05.Live translation
Recorded video is only half the catalogue; the other half airs live. Familiar translates streams in real time, each target language on its own channel running 10 to 30 seconds behind the original, at 42¢ per finished minute per language. It is the only voice + face translation in real-time: the stream stays your voice and your face while it airs in another language. Mechanics and setup are in Real-Time Dubbing for Livestreams.
06.Choosing by content type
- Talking-head videos. The face carries the message, so voice + face translation is the natural pick; subtitles waste the format.
- Podcasts and interviews. Multiple speakers must each keep their own voice; a single stranger voice-over flattens the conversation.
- Courses and lectures. Terminology is the risk: put technical terms on the Do Not Translate list and check them at the review screen.
- Sermons.A weekly recording reaching a multilingual congregation, in the speaker's own voice rather than an interpreter's.
- Livestreams. Only live dubbing works here; subtitles and recorded dubs arrive after the moment has passed.
- Ads and product video.Short runtimes make complete dubs cheap, and the founder's or spokesperson's real voice carries the trust.
Direction matters as much as method. Familiar's 25 languages run any-to-any, so a Spanish or Hindi creator can translate a video to English, the biggest audience unlock there is, just as easily as an English creator translates outward. Paired before-and-after examples are at /demos.
REFERENCES
Dubbing is finally good. See the measurements, then try it on your own video.