MediaPipe GenAI

Deploy Large Language Models (LLMs) and diffusion models for text and image generation directly on-device, supporting Android, iOS, Web, and desktop platforms.

MediaPipe GenAI delivers powerful, on-device generative AI capabilities via cross-platform APIs (Android, iOS, Web, Python). The suite includes the LLM Inference API for text generation, supporting models like Gemma-2 2B and multimodal Gemma-3n, and the Image Generator for diffusion-based image creation. Developers leverage key features: Retrieval Augmented Generation (RAG), Function Calling, and Low-Rank Adaptation (LoRA) for model customization. This enables fast, private inference on compatible devices (e.g., Pixel 7, iPhone 13, or newer).

https://developers.google.com/mediapipe/solutions