smolvlm-realtime-webcam
Framework for running vision models locally on an M3 MacBook — on-device vision demo
A framework for running vision models locally on an M3 MacBook — an on-device vision demo.
Image generation and computer vision — detection, editing, 3D.
Key features
- Simple demo of llama.cpp server with SmolVLM 500M for real-time object detection
- Runs a local web page (index.html) against the llama-server multimodal endpoint
- Customizable instruction, e.g. make the model return JSON
- Supports GPU acceleration via llama-server -ngl flag
Good for
- Experimenting with real-time webcam object detection locally
- Demo of on-device vision with llama.cpp
Stack & tags
JavaScriptHTMLvlmwebcamobject-detectionrealtimellama.cppsmolvlmdemo
Quick start
llama-server -hf ggml-org/SmolVLM-500M-Instruct-GGUFFrom the project README on GitHub.
Similar open-source projects
FLUX.2 klein
★ 25,996Black Forest Labs' sub-second image generation — generate and edit in <0.5s
DiffSynth-Studio
★ 13,200The framework behind Qwen-Image-i2L — the first "Image→LoRA" visual generation system