AI Tutorial Video Packager
Purpose
Turn a raw screen recording into a publishable tutorial video: create guide-style narration, synthesize Chinese voiceover, generate subtitles, prepend the fixed opener when available, and merge everything into one final MP4.
Quick Start
python <CODEX_HOME>\skills\ai-tutorial-video-packager\scripts\package_tutorial_video.py --input screen_recording.mp4 --title "Codex + AutoCAD" --output-dir outputs\published_codex_autocad
The script defaults to the local Windows WinRT Chinese male voice when available, keeps Edge neural TTS and SAPI as fallbacks, and uses imageio_ffmpeg for muxing/subtitle burn-in. It does not require ffmpeg on PATH.
Workflow
- Inspect the recording filename, duration, visible content, logs, and user prompt.
- Write a short guide-style narration focused on what the viewer learns, not only what happens on screen.
- Pass
--title, --topic, and up to three --highlight values. Use --script-file when Codex has written a custom narration.
- Let the script generate:
narration.txt
voiceover.wav
subtitles.ass
- optional
intro.mp4 and video_with_intro.mp4
final_upload_ready.mp4
- Verify the final MP4 has video, voiceover audio, and burned-in subtitles.
Narration Style
Use Chinese, practical, and guide-oriented wording:
- Start with the exact tool combination, such as
Codex + AutoCAD.
- Highlight the result early: Codex planning and AutoCAD automatic drawing.
- Explain the process as steps: prompt, planning, scripting, tool control, drawing, validation, output.
- Keep sentences short enough for subtitles.
- Avoid vague hype; use specific benefits and workflow tips.
Common Options
--input: source recording.
--title: opener title and narration topic.
--topic: more detailed subject; defaults to title.
--highlight: opener highlight line. Pass up to three times.
--script-file: custom narration text.
--output-dir: folder for all generated assets.
--tts-engine: winrt by default for local Chinese male voice; edge and sapi are fallbacks.
--winrt-voice: local Windows voice hint, default Kangkang.
--edge-voice: Edge neural voice, default zh-CN-YunxiNeural.
--voice-rate: Windows SAPI fallback rate, default 0.
--no-intro: skip fixed opener.
Coordination With Intro Skill
When ai-tool-video-intro is installed, use its scripts/make_intro.py to keep the fixed visual identity. Pass the actual episode title and highlights instead of generic text.
1---2name: ai-tutorial-video-packager3description: Package AI tool tutorial screen recordings into upload-ready self-media videos with narration scripts, Chinese voiceover, styled subtitles, fixed AI/traditional-tool intros, and final MP4 audio-video merging. Use for Codex plus AutoCAD, AI plus Office, CAD/CAM, coding, automation, workflow guide, tutorial, and strategy videos that need explanation, voice, captions, and publishing polish.4---56# AI Tutorial Video Packager78## Purpose910Turn a raw screen recording into a publishable tutorial video: create guide-style narration, synthesize Chinese voiceover, generate subtitles, prepend the fixed opener when available, and merge everything into one final MP4.1112## Quick Start1314```powershell15python <CODEX_HOME>\skills\ai-tutorial-video-packager\scripts\package_tutorial_video.py --input screen_recording.mp4 --title "Codex + AutoCAD" --output-dir outputs\published_codex_autocad16```1718The script defaults to the local Windows WinRT Chinese male voice when available, keeps Edge neural TTS and SAPI as fallbacks, and uses `imageio_ffmpeg` for muxing/subtitle burn-in. It does not require `ffmpeg` on PATH.1920## Workflow21221. Inspect the recording filename, duration, visible content, logs, and user prompt.232. Write a short guide-style narration focused on what the viewer learns, not only what happens on screen.243. Pass `--title`, `--topic`, and up to three `--highlight` values. Use `--script-file` when Codex has written a custom narration.254. Let the script generate:26 - `narration.txt`27 - `voiceover.wav`28 - `subtitles.ass`29 - optional `intro.mp4` and `video_with_intro.mp4`30 - `final_upload_ready.mp4`315. Verify the final MP4 has video, voiceover audio, and burned-in subtitles.3233## Narration Style3435Use Chinese, practical, and guide-oriented wording:3637- Start with the exact tool combination, such as `Codex + AutoCAD`.38- Highlight the result early: Codex planning and AutoCAD automatic drawing.39- Explain the process as steps: prompt, planning, scripting, tool control, drawing, validation, output.40- Keep sentences short enough for subtitles.41- Avoid vague hype; use specific benefits and workflow tips.4243## Common Options4445- `--input`: source recording.46- `--title`: opener title and narration topic.47- `--topic`: more detailed subject; defaults to title.48- `--highlight`: opener highlight line. Pass up to three times.49- `--script-file`: custom narration text.50- `--output-dir`: folder for all generated assets.51- `--tts-engine`: `winrt` by default for local Chinese male voice; `edge` and `sapi` are fallbacks.52- `--winrt-voice`: local Windows voice hint, default `Kangkang`.53- `--edge-voice`: Edge neural voice, default `zh-CN-YunxiNeural`.54- `--voice-rate`: Windows SAPI fallback rate, default `0`.55- `--no-intro`: skip fixed opener.5657## Coordination With Intro Skill5859When `ai-tool-video-intro` is installed, use its `scripts/make_intro.py` to keep the fixed visual identity. Pass the actual episode title and highlights instead of generic text.