Short Translate
Turn one take into another language with your own voice, with or without lip-sync, after a cost estimate.
short-translate takes one of your takes and gives it back in another
language, in your own voice. With HeyGen the lips follow the new language too.
Then short-editing edits the translated take like any
other, with captions transcribed in the new language.
It translates the take, not the finished short, so the editor cuts and times everything on the translated voice. It also cleans the take first: stutters and retakes sent to HeyGen would be translated, lip-synced and billed.
Invoke
/short-translate my latest take in English
/short-translate short-my-slug to Spanish
/short-translate <take> --to en --from fr
It always shows the duration and the cost first, and sends nothing until you say yes.
What it does
- Transcribes the take with word timestamps. A teleprompter take written
with
studio-scriptalready has its script next to it: the translation of each line is filled in for you. - Builds the cues: one per spoken sentence, with a natural translation (not word for word, names, numbers and product names kept).
- Cuts the stumbles: repeated words, false starts, a sentence said twice (the last good one stays). Pauses are tightened, each cut has a short fade. You see the kept length and the dropped ranges. It also writes a subtitle file timed on the cleaned take.
- Estimates, then sends after your yes. It runs 2 to 20 minutes in the background, and a job can be resumed if it gets interrupted, without being billed twice.
- Hands the result to the editor. The translated short is named
<original slug>-<lang>so both versions sit side by side. When the original short has a plan (scenes, B-roll, music), it is reused and retimed on the translated take instead of starting from scratch. On-screen texts and the CTA are rewritten in the new language.
Providers and modes
| Provider | What you get | Cost |
|---|---|---|
HeyGen precision (default) | voice and best lip-sync, side views, hands in front of the face | about 10 credits per minute |
HeyGen speed | voice and lip-sync, face mostly front-on | about 5 credits per minute |
HeyGen audio | your voice, original image, no lip-sync | about 4 credits per minute |
| ElevenLabs | your cloned voice, original image, no lip-sync | ElevenLabs credits |
Lip-sync only matters when your face is on screen for long stretches. A short full of B-roll or with a small webcam circle looks fine without it, and costs 2 to 2.5 times less.
What you need
HEYGEN_API_KEYin.short-editing/.envfor HeyGen (HeyGen, Settings, API), orELEVEN_KEYfor ElevenLabs (already needed by the kit).- Without the HeyGen key, ElevenLabs works in the meantime.
- HeyGen uploads are capped at 32 MB: a copy is re-encoded to fit. Beyond about 15 minutes, cut the take into parts first.
Next
- short-editing: edit the translated take.
- feedbacks: fix what is off after the first render.
- publish: put the translated version out on its account.