BaoCut turns video rough cuts into an agent skill
BaoCut just turned tedious video editing into single-line agent commands. Hooking directly into Claude Code or Codex, it automatically transcribes footage, strips filler words, and translates subtitles directly inside an editable timeline. Why it matters: Creators no longer need to scrub audio wave by wave just to fix a single bad take or subtitle line.

Anyone who edits talking-head videos knows the unique pain of scrubbing audio waveforms for hours just to delete pauses, fix bad takes, and sync subtitle timestamps.
BaoCut wants to wipe that entire manual chore off your desk. It hooks straight into Claude Code or Codex, allowing creators to run full transcription, cleanup, and rough cuts through plain text prompts.
Running editor commands through skills
While other tools try heavy MCP integrations with closed-source editors, BaoCut takes a much lighter approach. It wraps its native desktop CLI into an Agent Skill compatible with skills.sh.
Once hooked into an agent engine like Claude Code, a single command handles speech transcription, polish, and translation all at once. Need to clean up a video and export translated subtitles? You just tell the agent, and it triggers the background CLI parameters without you touching a mouse.
In practical use, it goes well beyond basic speech-to-text. It strips out filler words, detects repeated phrases, and cuts false starts cleanly on the timeline, without messing with the original source footage.
An editable timeline instead of a black box MP4
Most modern AI video tools suffer from the same fatal flaw: they take a prompt and hand you a finished, uneditable video. If word twelve in the translation sounds weird, your only option is to reroll the entire file from scratch.
BaoCut builds on an editable timeline where the project itself is the single source of truth. Subtitles, audio tracks, cuts, and translations all live in one synchronized workspace.
If you need to fix a single subtitle line or redo one synthesized sentence, the agent edits only that specific segment and issues a receipt of the change. You can drag clips by hand while the agent rewrites captions, sharing the exact same undo stack.
Under the hood, BaoCut runs local models like Whisper large-v3 or Qwen3-ASR on macOS and Windows, while also supporting cloud APIs, vocal separation, and frame stabilization.
Why it matters
For solo creators, cross-border content teams, and anyone repurposing foreign speeches, BaoCut turns hours of robotic subtitle alignment into a thirty-second agent prompt.
We think its real breakthrough is recognizing that creators do not want a black-box generator; they want a hyper-competent assistant inside an editable timeline. By keeping local edits local and maintaining human control, BaoCut solves real production friction instead of showing off vanity demos.
The only catch right now is setup. Because agent capabilities rely on developer environments like Codex CLI or Claude Code, non-technical creators might still face a learning curve before getting everything running smoothly.
Sources (1)
- github.com(opens in a new tab)Originalgithub.com