Views on Autopilot · Chapter 4B
Running on Mac and Windows
The book and library document both operating systems. Mac is the default path I use daily; Windows has a full Chapter 4B + this guide so you are not stuck.
Mac
MLX Chatterbox
PDF main chapters · factory ./start.command
Windows
Chapter 4B voice
Coqui XTTS / edge-tts · factory node runner.mjs
What stays the same
- Node +
npx hyperframes(compose / render) - ffmpeg polish chain (Chapter 5)
- Whisper captions via HyperFrames
- Output: mono 24 kHz WAVs into the composition
- Factory:
node runner.mjs→http://localhost:4300
Install shared tools (Windows)
- Node.js LTS from nodejs.org
- ffmpeg —
winget install ffmpegor static build on PATH - Python 3.11+ (add to PATH)
- Project folder e.g.
C:\Users\YOU\video-pipeline\
Path A — local clone (recommended)
Coqui XTTS v2 (PyTorch CPU or NVIDIA CUDA). Closest free “my voice from a reference WAV” option on Windows.
cd C:\Users\YOU\video-pipeline python -m venv .voice-clone-venv .\.voice-clone-venv\Scripts\activate pip install --upgrade pip pip install TTS torch torchaudio
Generate each line to WAV, then run the Chapter 5 ffmpeg chain. Full snippet: PDF Chapter 4B.
Path B — free non-clone (fallback)
pip install edge-tts edge-tts --voice en-US-GuyNeural --text "Your line." --write-media line01.mp3 ffmpeg -y -i line01.mp3 -ac 1 -ar 24000 line01_raw.wav # then Chapter 5 audio chain → mono 24 kHz polished WAV
Not your voice — still fine for volume / faceless channels.
Path C — hosted clone
ElevenLabs (or similar) if Path A is too heavy. Paid. Still download lines → same ffmpeg chain → HyperFrames.
Factory on Windows
cd path\to\video-factory node runner.mjs # open http://localhost:4300/dashboard.html
Point voice scripts at .\\.voice-clone-venv\\Scripts\\python.exe instead of Mac mlx paths. Zip download is in your library if you bought the ₹499 tier.