Social Good
Image to Tactile Graphic Converter
Convert diagrams into simplified tactile graphics ready for embossing
Advanced24h2–3
PythonOpenCVVision LLM
17/20Audio Description for Videos
Add spoken descriptions of on-screen action to videos.
Advanced24h2–3 people17/20 fit
Build Audio Description for Videos for blind and low-vision viewers in ~24h with 2–3 people. Stack: Vision LLM + FFmpeg + TTS.
Why it wins: Strong accessibility with multimodal AI.
Build flow
01 · Problem
Most online videos have no audio description.
02 · Solution
Detect scene changes, describe key frames with a vision model, fit descriptions into pauses and voice them.
03 · Build
- 1Extract frames
- 2Describe
- 3Align to gaps
- 4Mix audio
04 · Demo
Play a clip with generated descriptions.
MVP checklist
- Scene detection
- Frame descriptions
- Pause fitting
- Voice track
Tech stack
| Tech | Role | |
|---|---|---|
| Vision LLM | Image understanding | AI |
| FFmpeg | Tooling | Tool |
| TTS | Text-to-speech | AI |
Related ideas
More Social GoodSocial Good
Alt-Text Generator for Websites
Writes missing alt text so blind users can understand images
Beginner12h1–2Next.jsVision LLM16/20Social Good
NGO Paper Register Digitizer
Photograph paper registers and get a clean spreadsheet
Intermediate14h1–2Vision LLMNext.jsCSV17/20