Dialogue Splitter

Drop in a video. You get it back with the voices taken out, plus the voices on their own, ready to make a scene in Clip Studio.

Drop a video here Or choose a video MP4, MOV or WebM. Nothing is uploaded.

Then in Clip Studio

Unzip the download. Keep both files together in the same folder.
Start a new project from the video. The voices file is picked up on its own, because it has the same name.

Does my video get uploaded?
No. Everything happens in your browser, on your own computer. The video never leaves it.
Why is there a big download the first time?
The splitter is a voice-separation model of about 320 MB. It downloads once and stays saved in your browser, so after that it starts straight away.
What do I need?
A computer with a graphics card and an up-to-date Chrome or Edge. It doesn't work on phones. A faster graphics card means a faster split.
How long does it take?
It depends on your graphics card and the length of the video. The page shows how long is left while it works. Short scenes are quickest.
It removed some music too
Singing counts as a voice, so songs in a scene come out with the dialogue. Scenes with talking over background music and sound effects work best.
What's under the hood?
The BS-Roformer (viperx 12.97) voice model, running on your graphics card through WebGPU, with ffmpeg putting your video back together. The ffmpeg.wasm core is GPL-licensed; source details.