Install once, process locally
The bootstrap downloads the verified release, creates an isolated Python environment, installs FFmpeg when supported, and configures whisper.cpp. Media never leaves the machine.
Windows
Set-ExecutionPolicy -Scope Process Bypass -Force; irm https://av-proc.jezzick.net/install.ps1 | iexDetects NVIDIA, AMD, or Intel graphics. Uses CUDA transcription on NVIDIA, CPU transcription elsewhere, and prefers NVENC, AMF, or QSV video encoding before CPU fallback.
Linux or macOS
curl -fsSL https://av-proc.jezzick.net/install.sh | shSelects CUDA, ROCm, Metal, or CPU based on the available toolchain. Manual backend flags remain available after ZIP extraction.
One command per service
List camera files in chronological order and provide one continuous dedicated recording.
Supply the recordings
Use one --video for a normal service or repeat it for every interrupted camera segment.
Let the worker align
Transcript anchors establish coarse timing; waveform windows refine each segment and audio transition.
Publish only when ready
The result includes dual audio, subtitles, exact transition reports, and explicit release-gate status.
Built for imperfect recordings
- Freezes the last frame across genuine camera gaps.
- Removes duplicated overlap between consecutive clips.
- Uses attenuated scratch audio when dedicated audio is absent.
- Crossfades and limits the default listening track.
- Retains untouched camera scratch as audio track 2.
- Creates TXT, SRT, VTT, CSV, and JSON transcripts.
- Infers provisional titles from explicit transcript wording.
- Preserves reports and cached stages for review and resuming.