| Over the past few months, my wife and I have been working on a mobile app called FaceHarmony, which generates cinematic AI profile pictures. The app is now available on iOS and Android. What sets FaceHarmony apart is its unique approach to user control and customization. Many AI photo apps require users to pay for templated images without offering much say in the final result. To solve this, we’ve gamified the process: users swipe left on photos they don’t like and swipe right on the ones they want to download, similar to Tinder. This gives users more control over what they end up to pay for. Another personal aspect of this app is that after losing three close people in the past three years, I wanted a way to bring back their memories. FaceHarmony makes it easier to revisit and relive those moments by using lower-resolution input images (around 350x350) to create beautiful, lasting memories. I’ve been deeply involved in every aspect of this project, from design to launching managed instance groups on Google Cloud Platform, and I’m passionate about the technology, art direction and photography behind it. I’d love to hear your feedback. As a thank you, use the referral code HN20FREEPHOTOS during onboarding to create one model and download 20 photos for free. Happy to answer any questions! |
We’ve built FaceHarmony on Google Cloud Platform (GCP). We’re using SDXL to create those cinematic profile pictures, and managed instance groups with Google Batch to manage the heavy compute tasks for fine-tuning. It lets us scale easily based on demand, which is crucial for keeping everything running smoothly.
To make sure everything communicates well, we’ve set up Google Cloud Pub/Sub for messaging between our services. This setup helps us manage real-time interactions while the backend handles the AI processing.
For user authentication, we’re using Firebase Auth. We’ve also integrated Firebase Store Database to keep user preferences synced. On the backend, we’ve implemented FastAPI to build our APIs quickly and efficiently. We’re deploying these services on Google Cloud Run, which has been great for scaling our containers without worrying about server management. For the front-end, we’re using Vue 3 with TypeScript and Ionic to create a single codebase for both iOS and Android. This approach has helped us deliver a native-like experience on both platforms while keeping development time manageable. We’re using Google Cloud Storage to store user-generated content and model data.
I’m here if you have any questions or just want to chat about the tech!
Twitter: @OmidSakhiML LinkedIn: https://www.linkedin.com/in/omidsakhi/