InkWise

Snap a photo of any study material — textbook, notes, whiteboard — and InkWise instantly turns it into explanations, quizzes and flashcards using AI.

Comment

Inspiration

Honestly? Exam season. I kept staring at pages of notes not knowing if I actually understood them or just recognized the words. Existing tools either cost money or made you do all the work yourself. I wanted something that just got it from a photo.

What it does

You upload an image of any learning material a textbook page, handwritten notes, a whiteboard — and InkWise reads it using OCR, then generates a plain-language explanation, 6 quiz questions, and 8 flashcards automatically. You can also chat with the AI about the material and share your flashcard decks with a public link.

How we built it

Built the whole thing on MeDo which honestly surprised me with how far it got. Frontend is React + Tailwind with shadcn/ui components, backend runs on Supabase Edge Functions in Deno, and the AI pipeline chains OCR extraction into Gemini 2.5 Flash for content generation. The whole stack was described in natural language and MeDo scaffolded it out.

Challenges we ran into

Getting the multi-file upload to work correctly with a mix of JPEGs and PDFs in the same session was trickier than expected. Also the OCR → LLM pipeline needed careful prompting to consistently return structured JSON for quizzes and flashcards early attempts kept hallucinating extra fields or misformatting the response.

Accomplishments that we're proud of

The quiz and flashcard generation quality is genuinely good. Like it actually produces useful questions from real textbook content, not just surface-level stuff. And the whole thing was built without writing a single line of code manually which still feels a bit unreal.

What we learned

No-code doesn't mean no thinking. You still have to architect the data flow, design the prompts carefully, and know what you want the system to do MeDo just handles the implementation. Also learned that OCR quality varies a lot by image lighting which is something I want to fix next.

What's next for InkWise

Math rendering with KaTeX for STEM content, PDF multi-page support, and probably a mobile camera capture mode so you dont even need to leave the app to snap something.

Built With

  • deno
  • gemini-2.5-flash
  • medo-(no-code-platform)
  • ocr.space-api
  • react
  • shadcn/ui
  • supabase
  • supabase-(postgresql-+-edge-functions-+-auth)
  • tailwind-css
Share this project:

Updates