Genia, a mobile AI image generation app

Genia is a mobile AI image generation app, a personal project started in June 2026. The idea: a mobile image generation app in the spirit of Freepik or Magnific, where you describe what you want in French without knowing each model's quirks.

Gemini analyses the request and the reference images, then turns it into a structured JSON request: subject, framing, camera, lighting, palette, things to avoid. The user reviews this JSON, edits it if needed, and it becomes the prompt sent to the image model.

Every provider has its own rules: formats, resolutions, images per call, reference handling, rate limits. The app needed a single interface that only shows the settings the selected model supports.

Generation also takes time. On mobile the app goes to the background or gets closed, and an image that has already been paid for must not get lost. Finally, keeping API keys on the phone stops being acceptable as soon as the app goes beyond personal use.

I built the app with Expo SDK 56 and React Native, with a native Liquid Glass tab bar on iOS 26 and a blur fallback on Android. Keys are stored encrypted in SecureStore, and a demo mode lets you explore the app without any key.

I then added a Cloudflare Worker proxy that holds the keys and runs generations as durable jobs: the app polls their status and retrieves the images even after an interruption. Generations run in parallel with a cap per provider, and the latest versions add a style and character library, the exact cost before generating and push notifications.