Cinematic dubbing
Same voice.
Every language.
Upload a scene. Get it back in English, Hindi and Japanese, spoken by the same actor, with every breath and pause where they left it.
I never said I was leaving.
Wearebuildingdubbingthatkeepstheactor'svoice,breath,andperformanceacrosslanguages.ThecurrentfixedpreviewisEnglishtoHindi;Japaneseandmoredirectionsarestillbeingtested.
Every scene. Every language.
Example scenes
Monsoon Line
Drama·ENHIJA
Last Train to Shinjuku
Thriller·ENHIJA
Dust Road
Adventure·ENHIJA
The Quiet Court
Period·ENHIJA
Low Orbit
Sci-fi·ENHIJA
Sunday Lunch
Comedy·ENHIJA
Hear it
Switch the audio.
Keep the actor.
Dubbed by RelDub · Hindiतुमने कहा था तुम मेरा इंतज़ार करोगी।
Audio
- No account needed
- Consented files retained privately
- Demo output: evaluation only
- Consent confirmed for every voice
Upload. Walk away.
It finds you.
Upload a scene.
Up to 10 minutes of dialogue. Pick the language and confirm you have permission.
We dub it.
The model translates the lines and speaks them in the original voice. 5 to 10 minutes.
Check your inbox.
A private link to listen side by side and download. No waiting on the page.
What the dub keeps
Most tools translate the words.
We keep the actor.
Voice.
It still sounds like the person who said it, not a stock narrator.
Breath.
Sighs, laughs and the intake before a hard line stay in the take.
Pause.
Silence lands where the actor put it, so the scene keeps its rhythm.
Pitch.
Rising doubt, falling anger. The melody of the line carries over.
Timing.
The dub runs about as long as the source, so it still fits the picture.
The model
Our own model. Made for performances.
RelDub runs on a speech model we build in-house, for film and series dialogue, where how a line is said matters as much as what it says.
- Hears the performance
- It listens to how a line is delivered, not just what is said, and carries that delivery into the new language.
- Keeps the voice from a few seconds
- A short stretch of the speaker's own audio is enough for the dub to sound like them.
- Times itself to the picture
- Lines start where the actor starts and finish close to where they finish, so the dub sits on the cut.
- Speaks the way people talk
- Hindi comes out as everyday conversational Hindi, with English words kept where people actually use them.
- Any direction
- English, Hindi and Japanese, each into the others.
- Built to grow
- New languages are added on top of the existing ones without retraining them, so what works today keeps working.
Tech specs
- Languages
- English, Hindi and Japanese
- Directions
- English → Hindi fixed preview; more directions in testing
- Input
- MP3, WAV, M4A, AAC, FLAC, OGG or WebM
- Length
- Up to 10 minutes, 50 MB
- Turnaround
- 5 to 10 minutes, by email
- Storage
- Private; retained until removal request
- Training on your files
- Never
Behind the scenes
A new language.
In hours, not months.
Watch the model pick up a language it has never heard, from first listen to first dub.
क्या… तु… तुम
01
Listen
We give the model recorded speech in a language it has never heard.
02
Learn
A new language pack trains on top of the model. The languages it already speaks are left untouched.
03
Dub
The same voices, the same performances, now in one more language.
For studios and platforms
Bigger than one scene.
The model behind this page is built to sit inside a studio pipeline, a streaming platform or a localisation team. We're open to licensing, integration and deeper partnerships.
Catalogue scale
Batch dubbing through an API, from a single trailer to a full back catalogue.
Your infrastructure
Private-cloud or on-premise deployment when the footage can't leave the building.
Clean provenance
Every source the model learned from is licensed, synthetic or consented, and documented end to end.
Start your dub.
Clear dialogue works best: one speaker at a time, little background music.
Questions? Answers.
Which languages work?
The local preview currently plays one fixed English → Hindi sample. English, Hindi, Japanese are in audio evaluation, including both directions where clips exist. More languages are in research; no additional direction is available for public uploads yet.
Why do I have to wait?
The free preview runs on a GPU that may need to start first. Processing time varies, so we email you a private link when the result is ready.
Can I dub someone else's voice?
Only with their permission. You must own the recording or have the owner's written permission, and every speaker must have agreed to be dubbed. Dubbing someone to impersonate or deceive is against our terms.
What happens to my files?
With your permission, RelDub keeps the original and dub privately until you request removal. We do not train on your files. Your dub is for evaluation only: no commercial use or model training by you.
What audio works best?
Clear dialogue with one speaker at a time and little background music. Live uploads are still closed while we verify the worker and speaker matching.