Readability.Lib
Long stories, cut into two-minute reads with generated imagery.
People who grew up on short video rarely finish a novel. Readability.Lib does not argue with that habit; it works inside it. A story is broken into one- to two-minute passages, each paired with an illustration generated for that exact scene and small interactive beats that keep the reader moving.
The Django backend runs the whole pipeline: text segmentation, prompt construction, calls to Gemini and OpenAI for summarisation and scene description, GPT-2 for lightweight local generation, and Stable Diffusion for the artwork. Results are cached per passage so a book is only ever processed once.
It was selected as Engineering Project of the Year at Beykoz University in 2024.
- 2024
- Design, backend, AI pipeline
- Python, Django, Gemini, OpenAI, GPT-2, Stable Diffusion
- Engineering Project of the Year, Beykoz University
Plan a trip, find a travel mate, keep everything in one place.