Overview

A Bun-based CLI that processes recorded course videos by splitting chapter markers into separate files, trimming silence at the start/end, and optionally removing spoken "Jarvis" command windows via transcript timestamps refined with audio-based silence detection.

Requirements

• Bun - runtime and package manager
• ffmpeg + ffprobe - must be available on PATH
• whisper-cli (optional) - from whisper.cpp, required for transcription
  • Pass --whisper-binary-path if not on PATH
  • Model file auto-downloads to .cache/whispercpp/ggml-small.en.bin
• Silero VAD model - auto-downloads to .cache/silero-vad.onnx on first use

Installation

bun install

Quick Start

bun src/process-course-video.ts "/path/to/input.mp4" "/path/to/output" \
  --enable-transcription \
  --keep-intermediates \
  --write-logs

Web UI (experimental)

Start the Remix-powered UI shell in development mode (watch mode enabled):

bun run app:dev

To preload a local video path for the UI:

bun run app:dev -- --video-path "/path/to/video.mp4"

Then open http://localhost:3000. Use -- --port, -- --host, or -- --video-path to override the defaults.

For a production start without the file watcher:

bun run app:start

CLI Options

| Option | Alias | Description | Default | | ------------------------ | ----- | --------------------------------------- | ----------- | | input | | Input video file (mp4/mkv) | required | | outputDir | | Output directory | output | | --min-chapter-seconds | -m | Skip chapters shorter than this | 15 | | --dry-run | -d | Don't write files or run ffmpeg | false | | --keep-intermediates | -k | Keep .tmp files for debugging | false | | --write-logs | -l | Write log files for skips/fallbacks | false | | --enable-transcription | | Run whisper.cpp for command detection | false | | --whisper-model-path | | Path to whisper.cpp model file | auto-cached | | --whisper-language | | Language for whisper | en | | --whisper-binary-path | | Path to whisper-cli binary | system PATH | | --chapter | -c | Filter to specific chapters (see below) | all |

Chapter Selection

The --chapter flag supports flexible selection:

• Single: --chapter 4
• Range: --chapter 4-6
• Open range: --chapter 4-* (chapter 4 to end)
• Multiple: --chapter 4,6,9-12

Chapter numbers are 1-based by default.

Output Structure

Final files are written to the output directory with names like:

chapter-01-intro.mp4
chapter-02-getting-started.mp4
chapter-03-custom-title.mp4

When --keep-intermediates is enabled, intermediate files go to output/.tmp/:

| File Pattern | Description | | --------------------- | ------------------------------------------------ | | *-raw.mp4 | Raw chapter clip with initial padding removed | | *-normalized.mp4 | Audio normalized (highpass + denoise + loudnorm) | | *-transcribe.wav | Audio extracted for whisper | | *-transcribe.json | Whisper JSON output | | *-transcribe.txt | Whisper text output | | *-splice-*.mp4 | Segments before/after command windows | | *-spliced.mp4 | Concatenated output after command removal | | *.log | Per-chapter skip/fallback logs | | process-summary.log | Overall processing summary |

Voice Commands

Commands are spoken in the format: jarvis <command> ... thanks

| Command | Effect | | --------------------------------------- | ----------------------- | | jarvis bad take thanks | Skip the entire chapter | | jarvis filename my-custom-name thanks | Rename output file |

The command window (from "jarvis" to "thanks") is removed from the final video.

More Details

Implementation notes and pipeline details live in docs/README.md.

This project was created using bun init in bun v1.3.1. Bun is a fast all-in-one JavaScript runtime.