Quote-image generator that refuses to fake your subtitles
An Agent Skill that turns video frames into ready-to-post 3:4 quote cards — while keeping an unusually strict line between burned-in subtitles and text added after the fact.

What it does Feed it a local video or a YouTube link and it picks quotable lines, grabs the exact frames, and composes publishable 3:4 JPG quote cards — one hero frame plus stacked subtitle strips, with per-image quality checks. It ships as an open-format Agent Skill (Codex, Claude Code, others) but the underlying Python scripts run fine on their own.
The interesting bit The project’s whole personality is subtitle honesty. “Native” mode only crops subtitles that are physically burned into the pixels — no OCR, no redrawing, no translation. “Script” mode draws your reviewed lines onto real frames but explicitly labels them as post-production. The two never mix, and if you ask for native subtitles on a video that only has toggleable CC tracks, the agent tells you before switching modes.
Key highlights
- Two modes,
renderandrender-script, with the source of every word documented - Layout auto-adjusts: one hero image plus four subtitle strips fills ~70% height, gapless
samplecommand generates a candidate contact sheet so you stop guessing timestamps- Multi-image jobs also emit a
final_contact_sheet.jpgoverview for quick picking - Works fully offline for local files; YouTube ingestion is optional via
yt-dlp
Caveats
- Won’t upscale low-res footage into “real” HD — it says so explicitly
- Script mode requires CJK fonts for Chinese/Japanese/Korean lines; you may need to point it at one
- You’re on the hook for usage rights on the source video
Verdict Anyone making social quote cards from subtitled video — clip accounts, content teams — will save an hour per post. If your video only has soft subtitle tracks and you want “native” text, it will (correctly) tell you no.
Frequently asked
- What is chengyi-ai/native-subtitle-quote-image?
- An Agent Skill that turns video frames into ready-to-post 3:4 quote cards — while keeping an unusually strict line between burned-in subtitles and text added after the fact.
- Is native-subtitle-quote-image open source?
- Yes — chengyi-ai/native-subtitle-quote-image is open source, released under the MIT license.
- What language is native-subtitle-quote-image written in?
- chengyi-ai/native-subtitle-quote-image is primarily written in Python.
- How popular is native-subtitle-quote-image?
- chengyi-ai/native-subtitle-quote-image has 2.9k stars on GitHub.
- Where can I find native-subtitle-quote-image?
- chengyi-ai/native-subtitle-quote-image is on GitHub at https://github.com/chengyi-ai/native-subtitle-quote-image.