Loading the detector…
Vision Lab
Pick a photo or snap one with your webcam. An object detector runs right in your browser and boxes what it sees, then you can ask a vision model anything about the picture.
How it works
- On-device detection.
Xenova/yolos-tinyruns on Transformers.js in your browser (WebGPU when available, otherwise WebAssembly). The photo is never uploaded for detection. - Vision chat. For questions, the image is shrunk to 1024 px and sent with your question to a free multimodal model on OpenRouter through AI Valley's proxy.