Tool · Automation
Xưởng Tâm Sự (video automation)
Automatically edits Japanese YouTube videos: kanji subtitles synced to voice, music and motion. A 44-second video renders in 15 seconds.
- Role
- Concept · Design · Development
- Year
- 2026
- Tech stack
- Python · FastAPI · FFmpeg · Whisper · BudouX
IllustrationContext
A heart-to-heart YouTube channel for Japanese viewers needs a steady release schedule, but hand-editing each episode — subtitles, music, transitions — took far too long.
Challenge
Subtitles had to keep the script’s exact kanji yet match the AI voice word by word — and the output had to be good enough to publish as is.
Solution
Whisper provides word timestamps, then difflib aligns them with the original script to keep the kanji intact.
BudouX breaks Japanese subtitle lines at natural phrase boundaries.
FFmpeg builds 16:9 scenes with zoom, transitions and a dust overlay, and mixes voice, music and ambience.
Runs locally in its own window, listening on 127.0.0.1 only to keep API keys safe.
Results
15s
to render a 44-second video
1080p
H.264, ready to publish
89%
test coverage (45 tests)