Your AI agent for
subtitles & video.
Turn any video, audio, or recording into editable subtitles — then proofread, translate, segment, and cut. Transcribe, translate, cut, all on your Mac.
So I want to start with the very first time.
00:06 / 41:05So I want to start with the very first time you touched clay.
Do you remember it?
4 words · 1 cue · reads clean
Local-first & private
Transcription and speaker ID run entirely on your Mac. Your audio never leaves the device — cloud models are used only when you turn them on.
Edit like a document
Words are the single source of truth. Find and replace, undo and redo, press Enter to split a line, Backspace to merge. Cues and timing follow the text.
AI suggests, you decide
Every AI step — segment, polish, translate, cut — lands as a reviewable suggestion. Nothing changes your transcript until you accept it.
Transcribe · Translate · Cut
Three verbs, one clean workflow
From raw footage to polished, translated, cut subtitles — without leaving the editor or your Mac.
On-device speech to text, down to the word
Voice activity detection, ASR, and forced alignment run locally to produce word-level timestamps — so every cue lands exactly where the words are spoken.
- Qwen3-ASR or Whisper, running on Apple Silicon with MLX
- Speaker identification — who spoke, and when
- Word-level timing you can trust, not sentence guesses
The material teaches you patience before it teaches you anything else
Bilingual subtitles that actually fit
Translate line by line with the source and translation aligned sentence to sentence — reviewed, reading-speed aware, and broken the way each language reads.
- Source ↔ translation aligned per sentence, reviewed one pair at a time
- Characters-per-second (CPS) budgets keep lines readable
- Line breaks tuned to each language, not copied from the source
Edit the video by editing the text
Remove filler and retakes with soft, reversible cuts. Add B-roll over the talking head. Then export subtitles or a burned-in video — the cuts and captions always stay in sync.
- Soft cuts are reversible and cue-aligned — nothing is destroyed
- B-roll overlays, full-screen or picture-in-picture
- Export SRT, VTT, ASS, Markdown, or a captioned video
I think um, I think the honest answer is that it just felt right.
The Descript model
Subtitles you edit like text
Words are the only source of truth. Subtitle cues and paragraphs are just projections of the text — so editing reads like a document, and the timeline keeps up automatically.
Let's ship the new color system this week, then circle back on the docs. If the color tokens land clean, everything downstream just works
Private by default
Your audio stays on your Mac
The whole transcription pipeline runs on-device with Apple's MLX. Cloud models are strictly opt-in, and only for the steps you choose.
On your Mac
Video, audio, and recordings — imported locally, never uploaded.
Editable subtitles
On-device VAD, ASR, forced alignment, and speaker diarization.
Assist, never override
AI flows you stay in charge of
Each flow follows the same three beats — set up, run, review. The result is a suggestion you apply in one step, and can always undo.
Segment
Break the transcript into clean paragraphs the rest of the pipeline builds on.
Polish
Fix transcription slips and punctuation while preserving every word timing.
Chapters
Generate chapter markers from the content, ready to name and reorder.
Speakers
Re-identify and rename voices, then apply labels without touching the text.
Cues
Re-flow subtitle line breaks for reading speed and natural phrasing.
Translate
Produce aligned bilingual subtitles, reviewed pair by pair before they land.
In and out
Bring anything in, take standards out
Import
Export
Built for agents
Drive the whole pipeline from the command line
BaoCut ships a companion CLI that runs the same real engines headlessly — transcribe, polish, and translate in one command, or step each stage under an AI agent.
- One command, end to end:
transcribe → polish → translate - Export SRT, VTT, ASS, Markdown, or video from a project id
- The GUI and CLI share one project — open a run mid-flight
Download BaoCut for Mac
Transcribe, translate, and cut — locally, on your Mac. Free to use, no account required.