Inspiration
We noticed how often people freeze or stumble when trying to craft the perfect opening line or reply in a romantic chat. In today’s fast-paced dating world, confidence often comes down to having the right words at the right moment. We asked ourselves: what if AI could become your personal “wingman,” analyzing your chat context and generating spot-on flirty responses in an instant?
What it does
Flirt AI leverages on-device screenshot analysis and server-side language models to deliver:
- Subtle Openers when you want to break the ice.
- Witty Replies to keep the conversation flowing.
How we built it
- Platform & UI: Developed entirely in SwiftUI, using modern view composition and
async/awaitfor smooth image-to-AI flows. - AI Integration: Backend hosted on Cloudflare Workers, forwarding prompts to OpenAI’s GPT & Google Gemini
- Screenshot Analysis: We use VisionKit to securely extract text and context from user screenshots—no data is stored on-device.
- Monetization: Subscriptions and unlocks (e.g., “Unlimited Pick Lines,” ) are handled via RevenueCat, giving us unified cross-platform IAP management and real-time entitlement checks.
Challenges we ran into
- Privacy & Permissions: Ensuring users feel safe granting screenshot access; we implemented on-device text extraction and explicit consent flows.
- Subscription Edge Cases: Handling trial expirations, downgrades, and restore-purchases reliably across sandbox and production—RevenueCat’s webhooks and customer-info APIs were lifesavers.
- UI Performance: Keeping the SwiftUI views responsive while waiting for API calls required careful state management and lightweight placeholder views.
Accomplishments that we're proud of
- MVP in 48 Hours: From blank Xcode project to working prototype with screenshot upload, AI response, and live IAP gating.
- Seamless RevenueCat Integration: Unified trial, monthly, and yearly plans with no flicker or UX jank.
- Positive Early Feedback: Beta testers report 95% “would recommend” and cite the “one-tap charm boost” as game-changing.
What we learned
- SwiftUI’s Power: How to architect a modular, data-driven UI that stays snappy under asynchronous workloads.
- Prompt Engineering: Crafting context-aware templates that coax out playful, sincere, or bold tones from the same AI model.
- IAP Best Practices: RevenueCat drastically simplifies cross-platform subscriptions but still demands rigorous testing of restore and entitlement flows.
What's next for Flirt AI
- Tone Presets & Themes: “Movie-Quote Mode,” “Shakespearean Mode,” etc.
- Integration with Chat Apps: One-tap sharing into WhatsApp, iMessage, and beyond.
- Android Companion App: Bringing the same seamless experience to a wider audience.
Log in or sign up for Devpost to join the conversation.