Most streak apps count a tap. You can keep a streak alive with a blurry photo, an old picture or a screenshot. StreakSphere fixes fake streaks with AI-verified proof: a daily activity only counts after an AI has looked at it.
On StreakSphere, a streak only grows when an AI has checked the proof. No real activity, no points.
How a proof is checked
- You capture it in the app. A photo or a short video of the activity: a workout, a run, a study session, a healthy meal, a walk outside.
- Blank or covered shots are stopped first. Dark, blank or one-colour images and videos are rejected straight away, without counting as a violation.
- Re-used proofs are caught. Every proof gets a fingerprint of the photo. If it looks the same as one of your verified proofs from the last 30 days, it is rejected as a duplicate, so an old photo can't be posted twice.
- A safety check runs before anything else. The AI service first screens the image (for a video: three frames spread across it) with an image-safety model. Explicit content is blocked and counts as a strike; borderline content is marked age-restricted.
- The AI matches the photo with real activities. OpenAI's CLIP model compares the picture with descriptions of real activities, such as "a person working out at the gym", "a person doing yoga on a mat" or "a healthy plate of food". If the photo clearly shows one of them, the proof is verified.
- Verified proof moves you forward. It earns points (twice as much with StreakSphere+), advances your streak, and counts on the leaderboards for your city, your country, the world and your friends. A rejected proof earns almost nothing; blank, re-used and unsafe proofs come with a notification explaining why.
Why it matters
Leaderboards and streaks only mean something when they can't be gamed. Because every point on StreakSphere comes from a checked proof, a top spot in your city is earned by showing up, not by tapping a button. The same idea runs through the whole app: moods are live, letters are private, and effort is real.
The technology behind it
- App: React Native, with the in-app camera for proofs.
- Backend: Node.js (Express) and MongoDB; a background worker queues each proof and runs the blank, duplicate and AI checks.
- AI service: Python with FastAPI, OpenAI's CLIP (ViT-B/32) for activity matching, a Hugging Face image-safety classifier and OpenCV for video frames.
StreakSphere is free on Google Play. More about StreakSphere and who built it.