Mog.GPT is a browser-based AI system that quantifies visual dominance in group images. Instead of relying on subjective impressions, the platform analyzes measurable signals — spatial presence, posture dominance, facial intensity, and compositional positioning — to generate comparative dominance scores for each individual.
The system combines multiple client-side computer vision models with generative AI to produce both structured metrics and natural-language explanations. Users can also interact with a conversational voice agent powered by ElevenLabs to discuss their results in real time.
All heavy computation runs directly in the browser using TensorFlow.js and pose detection models, supported by performance optimizations like Web Workers and lazy loading. Mog.GPT demonstrates how multi-model AI orchestration, real-time visualization, and conversational interfaces can transform abstract human traits into interpretable, interactive insights.
Built With
- claude
- elevenlabs
- tensorflow
Log in or sign up for Devpost to join the conversation.