Local-first
Projects, media, and models stay on your computer. Cloud models run only when you choose them.
Tell the agent what you want, or do it by hand. BaoCut 2 transcribes on your computer, translates subtitles that fit the screen, and dubs your video into another language in the speaker’s own voice.
Describe the video you want in one sentence. The agent inside BaoCut runs the real workflow — fetching the video, transcribing, translating, aligning, and checking — and reports back in the same thread.
Bring in a video, an audio file, or a link. Speech is recognized on-device, lines appear as they are heard, and every word keeps its own timestamp.
Translate sentence by sentence, then read the source and the translation side by side. Each translated line is timed to the words it says, broken where it reads naturally, and checked for reading speed.
Turn the translation into speech. BaoCut can clone each speaker so the dub still sounds like them, fit the translation to the original timing, and generate the voices on your computer.
And the rest of the editor
Projects, media, and models stay on your computer. Cloud models run only when you choose them.
Split and arrange video, audio, text, images, and elements on one timeline.
Find filler words, long pauses, and retakes, then choose what to cut.
Ask for a product intro, an explainer, or a whiteboard or math lesson. No footage needed.
SRT, VTT, ASS, Markdown transcripts, or MP4 with the subtitles burned in.
The bcut command line and a built-in MCP server let other agents drive the same engine.
Install the CLIOpen source
Install the BaoCut skill, and Claude Code, Codex, or any skills.sh agent can transcribe, translate, subtitle, and cut through the bcut command line — you just ask in plain words.
BaoCut.app is optional. The skill uses bcut from your PATH, or downloads a verified, release-pinned copy. Installing with skills.sh needs Node.js.
Previous version
The original BaoCut app, version 1.1.4. It keeps its own update channel and needs macOS 15 or later.