I Turned Raspberry Pi into a Talking AI Camera

Let's run Hugging Face’s SmolVLM, the world smallest vision-language models on Raspberry Pi 5, with the web camera input and piper-tts output. It wasn’t as simple as I hoped—between slow inference times with onnx runtime and missing support in llama.cpp, I had to dig through GitHub issues and hack together a solution. In this video, I’ll show you how I got it working, the challenges I faced, and how you can try it yourself! If you in need of a Raspberry Pi, get it here: Raspberry Pi 5, Options for 2GB/4GB/8GB RAM https://s.click.aliexpress.com/e/_ExScjyg Github repository: Multi-modal support is back to llama.cpp! Updated script here https://github.com/AIWintermuteAI/pi-talking-camera Llama.cpp issues and PRs https://github.com/ggml-org/llama.cpp/pull/11292 https://github.com/ggml-org/llama.cpp/issues/8010