Local clip studio · phase one
Find the moments
worth posting.
Drop in a long video. Tiny Whisper listens locally, then Bougie shapes the strongest moments into editable short clips.
◆ Your video stays on this device. English sources benchmark local transcription first; if a cloud route is needed, only an audio-only copy is sent securely through our server. Timestamped transcripts and your instructions are then sent through our server for clip analysis.
Confirm your speakers
Match each voice to a face.
Deepgram found speaker turns. Confirm the people in this source before Bougie plans jump cuts.
Each play control shows the original source video while that Deepgram-labeled voice is speaking. This confirmation is saved for the project. If an edited source cuts the active voice out of frame, Bougie follows the live visible source person or a safe source crop—it does not freeze, retime, or invent the absent speaker.
Getting the local model ready.
You can work in other tabs or apps. Keep this tab open and your computer awake. Browser power saving may pause processing.
Choose local Tiny Whisper to transcribe English on this device, or Deepgram for cloud transcription and diarization. Deepgram receives an audio-only copy; your source video stays on this device.
Auto uses local English transcription only within its duration and measured-speed envelope. Selecting Local Tiny Whisper bypasses Auto's admission checks, but it still falls back safely if pure inference reaches 6:00.
- Source
- Awaiting source
- Project storage
- Awaiting a source · master files stay on-device
- Transcription route
- Waiting for source duration
- Model
- Tiny Whisper · q8 · WASM
- Audio preparation
- Idle
- Transcription
- Idle
- Route diagnostics
- No completed transcription run
- Transcript intelligence
- Waiting for transcript
- Speaker reframing
- Waiting for clip ranges
- Current visual decision
- Waiting for visual evidence
- Rendering
- Waiting for approved clips
- Awaiting adaptive transcription events
- Completed steps
- —
Your selected route, actual route, fallback reason, timing, transcript version, and cache use appear here. This panel never displays transcript text, signed URLs, or provider credentials.
Your cut room
0 moments, ready to shape.
With captions · default style. Outputs save to your account; download either version from its card.
Correct the words
Only this clip’s captions change. Click a timestamp to listen. Replacing words keeps their timing; adding or removing words estimates timing within that line.
You can work in other tabs or apps. Keep this tab open and your computer awake. Browser power saving may pause processing.
Want to compare transcription routes? Choose one below and explicitly rerun this project. Your current result is retained until the replacement completes.
- Source
- Awaiting source
- Project storage
- Awaiting a source · master files stay on-device
- Transcription route
- Waiting for source duration
- Model
- Tiny Whisper · q8 · WASM
- Audio preparation
- Idle
- Transcription
- Idle
- Route diagnostics
- No completed transcription run
- Transcript intelligence
- Waiting for transcript
- Speaker reframing
- Waiting for clip ranges
- Current visual decision
- Waiting for visual evidence
- Rendering
- Waiting for approved clips
- Awaiting adaptive transcription events
- Awaiting first run
- —
Route and timing details for this project. Transcript text, signed URLs, and provider credentials are never shown here.
Add a custom clip Your timestamps
Render with the default captions, or save an uncaptioned version for your own styling. Both versions keep your framing and audio and save privately to your account.