Dubbing a clip into another language on LUVI: four steps and what each one costs

Putting a clip into a second language is four separate jobs on LUVI, and only one of them is expensive. You pull the words out, rewrite them, have them spoken, and press the new voice back onto the mouth. On September 22, 2026 we quoted each step on a real ten-second clip: transcription came to 4 credits a minute of audio, voicing ran from 30 credits per thousand characters, and the two lip-sync models came to 130 and 2,200 credits for the same ten seconds. Here is the chain, and where the money actually goes.

Step one: get the words out

Transcription is the cheapest thing in this workflow. Grok's stt-v1 is billed per minute of audio at 4 credits a minute, so thirty seconds of speech costs 2 credits to transcribe—rounding included.

The transcriber listens to audio, not video. Its one input slot accepts an audio file only, so what goes in is the clip's soundtrack as an MP3, WAV, OGG, AAC or M4A file, not the clip itself.

The transcript arrives in your library as a text result rather than an audio file, which is worth knowing because it behaves differently from everything else you generate: transcripts are deliberately excluded from the reference pickers, since a transcript is never a model input.

Step two: rewrite, don't translate

This step costs nothing and matters most. A dub is not a translation with the same word count—it is a rewrite that fits the mouth it has to sit on. Two constraints shape it:

• Length. The new line has to run about as long as the old one, or the lip-sync model will either run out of mouth or run out of audio. Count syllables, not words.
• Stress. Put the emphasis where the face already puts it. If the speaker's eyebrows go up on the third beat, the new line needs something worth emphasizing there.

Read the transcript, write the new line to the clock, and only then move on. This is the step where a dub is won or lost, and it happens outside the models entirely.

Step three: voice the new line

Any of LUVI's text-to-speech models will do it, and the rate spread is wide: from 30 credits per thousand characters on Grok Speech up to 300 on the newest Gemini tier, with ElevenLabs v3 at 200. Thirty seconds of speech is roughly 450 characters, which works out at about 14 credits on the cheapest model and 90 on ElevenLabs.

Match the voice to the face before you match it to your taste. A clip of a woman in her sixties needs a voice that could belong to her; a voice that sounds twenty-five will fight the picture no matter how good the lip-sync is.

Step four: put the voice back on the mouth

This is the step that decides the bill, and the two models on LUVI differ by more than an order of magnitude. On the same ten-second clip, VEED Lipsync quoted 130 credits and Sync Lipsync v3 quoted 2,200. That is 13 credits per second of video against 220.

Both take the same two inputs—a video and an audio file—and neither takes a prompt at all. VEED's model has no parameters whatsoever, which makes it the simplest model in the catalog to drive.

What does the whole thing cost?

For a thirty-second clip, using the arithmetic above:

• Transcribe: 2 credits
• Rewrite: free
• Voice, on the cheapest model: about 14 credits
• Lip-sync on VEED: 390 credits

That is around 406 credits for a thirty-second dubbed clip. Swapping in ElevenLabs for the voice takes it to about 482. Swapping in Sync Lipsync v3 instead takes it past 6,600, which is the number to keep in mind when you are deciding whether the extra fidelity is what this particular clip needs.

How we checked

• Models: VEED Lipsync and Sync Lipsync v3, quoted against a real ten-second clip; Grok stt-v1, Grok Speech tts-v1 and ElevenLabs v3 for the transcription and voicing rates.
• Settings: a 1280×720 clip of ten seconds; default settings on every model.
• Date and what was read: September 22, 2026—live credit quotes and the models' parameter contracts in LUVI.
• Results: 130 credits on VEED Lipsync and 2,200 on Sync Lipsync v3 for the same ten seconds; 4 credits per minute for transcription; 30 and 200 credits per thousand characters for the two voice models named. The thirty-second totals above are arithmetic on those per-second and per-character rates, not separate quotes.
• Drawbacks: no outputs were generated for this post, so it makes no claim about which lip-sync model looks better, or about how well either handles a language whose mouth shapes differ from the original.

In LUVI

Upload the clip and its soundtrack as a separate audio file, run the transcription on the audio, and keep the transcript open beside you while you write the new line—the library shows text results inline. Then generate the voice, and feed the clip and the new audio to the lip-sync model as its two references; the reference media guide covers how to attach both. If you are doing several clips, the workflow canvas lets you wire the steps once and run them per clip instead of doing it by hand each time.

New accounts start with 200 credits, and verifying a phone and a card adds up to 1,500 more. You can create a LUVI account and try the chain on one clip before committing a series to it.