context-to-video

Context → Video

Convert any input context into an mp4 with narration + slides + subtitles + (optionally) an avatar.

When to Use

動画化, 解説動画作って, ブログを動画に, 記事を動画化, explainer video, context to video
Want to turn release blogs / PR descriptions / meeting minutes / design memos / release notes / learning content into short videos
Want to mass-produce materials for YouTube / internal study sessions / Teams / X in less time
If distributing to customers, replace TTS with the official Azure Speech version (default is the free edge-tts endpoint)

Input Sources (any of these works)

Input	How to obtain
Blog/article URL	`web_fetch` fetch the body with
Pasted text	Summarize as-is
PR / commit diff	`gh pr view` + `gh pr diff`
Meeting minutes/design memo	Read file
Release Notes	URL or Markdown
Free prompt	Ask the LLM to structure it directly

→ Processing after source acquisition is common (the LLM writes notes/script.json ).

Stack (all free, local)

Role	Tools
Script generation	LLM (the caller of this skill)
Slides	Pillow (BIZ UDGothic, etc.)
Narration	`edge-tts` (Edge Neural voices, free, unofficial)
Video composition	`ffmpeg` (`imageio-ffmpeg` bundled binary allowed)
Subtitles	`.srt` auto-generated
Avatar (optional)	SadTalker (GPU recommended) + ffmpeg elliptical mask

Quick Workflow (no avatar)

pip install edge-tts imageio-ffmpeg pillow
# 1. 入力からscript.json生成 (LLM)
# 2. 動画生成
python scripts\build_video.py --project <project-dir>
# → output\final.mp4 + final.srt

Details: references/script-schema.md / references/setup.md

Avatar Mode (talking-head in corner)

If you have a GPU (VRAM 8GB+), you can add an avatar in the bottom right with SadTalker. Initial setup takes 30–40 minutes in a separate venv (Python 3.10), and thereafter +20–30 minutes per video.

# 既存動画 + 8音声 → アバター合成
python scripts\add_avatar.py --project <project-dir> --face <face.jpg>
# 楕円マスク版を作る
python scripts\ellipse_overlay.py --project <project-dir>
# → output\final_with_avatar.mp4 と final_with_avatar_ellipse.mp4

Setup, pitfalls, character selection: references/avatar.md

Output Layout

<project>/
├── notes/script.json                  台本 (LLM が書く)
├── slides/slide_XX.png                1920x1080
├── audio/audio_XX.mp3                 スライドごとのナレ
├── avatar/                            (任意)
│   ├── input/face.jpg                 アバター元写真
│   ├── SadTalker/                     repo + checkpoints
│   ├── venv/                          Python 3.10 + torch cu121
│   └── clips/audio_XX.mp4             各ナレぶんアバター動画
└── output/
    ├── final.mp4                      無印 (スライド+ナレ)
    ├── final.srt                      字幕
    ├── final_with_avatar.mp4          (任意) アバター矩形
    └── final_with_avatar_ellipse.mp4  (任意) アバター円形ワイプ

Defaults

Resolution: 1920x1080 30fps (16:9)
Voice: ja-JP-NanamiNeural (female, calm). For male, ja-JP-KeitaNeural
Number of slides: cover + 5–7 main slides (6–8 total)
Color scheme: slate-900 / sky-400 / BIZ UDGothic
15–25 seconds per slide, 2–4 minutes total

Changing voice, color scheme, font, aspect ratio: references/customization.md

Production Use

Free edge-tts は unofficial, no SLA, commercially gray-area. For customer distribution or continuous operation:

Narration → Azure Speech (Neural TTS) replace with
Commercial with avatar → Azure TTS Avatar (Lisa, etc., official, from $0.30/min)

Details: references/production-upgrade.md

Common Variants

9:16 short version (X/Reels): W,H = 1080, 1920, compress narration to 60 seconds
Burn in subtitles: ffmpeg subtitles=final.srt burn into the video with the filter (ffmpeg-advanced-filters.md)
Add BGM: mix a separate mp3 at low volume (ffmpeg-advanced-filters.md)
Chapter/course format (learning material): split script.json by chapter, common intro/outro
Multilingual version: voice switch + translate the script
Channel-fixed character generation: create a mascot with SDXL and save the seed (avatar-generation-sdxl.md)
VOICEVOX + official character standing art: for "relaxed/friendly explanation" channels like Zundamon, full lip-sync with PSD layer differences (voiced-mascot.md)
Circular avatar wipe: circular mask + accent ring with ffmpeg geq filter (ffmpeg-advanced-filters.md)
YouTube auto publishing: public scheduling + thumbnail via API (youtube-upload.md)
PoC → promotion to regular streaming channel: directory redesign (scale-up-production.md)

Done Criteria

<out>/output/final.mp4 (or avatar version) can be played, and audio and video are not out of sync
Slide body text 28pt or larger / title 56pt or larger is readable
Narration correctly reflects the main message of the input
final.srt is synchronized with the narration
Does not infringe the copyright of the input source (summary + quotation range, not reading the full text)
(when using avatar) completes without GPU memory overflow, and lip-sync is not extremely off