LookSharp
AI-assisted build (vibe coded) using Codex and Cursor. Built through iterative prompting and refinement; AI tools contributed to the implementation.
The problem
Trying eyewear online needs a useful preview without losing the identity and context of the original portrait.
How I solved it
Connected Gemini 2.5 Flash Image to a Next.js try-on workflow with catalog, upload and URL references. Validated images and saved generated results with model/prompt metadata in a browser gallery.
- Image validation and try-on endpoint
- Persistent generated-image gallery
The results
- eyewear reference inputs
- 3 paths
- default image upload limit
- 10 MB
Delivered three eyewear reference paths: catalog selection, custom upload and image URL. Added a 10 MB default upload cap, generation timing and an IndexedDB gallery with save/delete operations.
Evaluation context. Implemented scope and configured limits. No measured conversion uplift, average generation latency or adoption claim.