New Video Generator | Generated by AI

Home 2026.07 English 中文

---

All done. Here’s a summary of what was created:

ww gen-video <file> — generates a vertical short-form video (9:16) from a markdown note.

Files changed:

• ww/gen_video/__init__.py — empty package init
• ww/gen_video/video.py — main module (~680 lines)
• ww/main.py — help text + dispatch block + group lookup
• ~/.zsh/completions/_ww — zsh completion entry

Pipeline:

1. LLM script generation — reads markdown, uses your default MODEL to create a narration script + 5-6 image prompts per scene
2. Flux image generation — calls black-forest-labs/flux.2-pro via OpenRouter for each scene, handles base64 data URLs and brand-name moderation
3. TTS narration — uses macOS say command (Samantha/Tingting voice), converts to AAC
4. FFmpeg assembly — stitches images into a 1080×1920 vertical video with narration audio

Usage:

ww gen-video /path/to/note.md
ww gen-video /path/to/note.md --output my_video.mp4 --voice Samantha
ww gen-video /path/to/note.md --audio existing_narration.aac
ww gen-video /path/to/note.md --image-model black-forest-labs/flux.2-pro

Tested with the Tesla P100 vs M60 article — produced a 164s, 3.9 MB, 1080×1920 video with 5 Flux-generated images and narration.

---

Generated by AI. Curating and sharing still takes effort. If you find it useful, feel free to donate. WeChat: @lzwjava · X: @lzwjava · Say hi 👋

Back Donate