MediaPipe On-Device Text-to-Image Generation Solution Now Available for Android Developers

Earlier this year, MediaPipe Solutions introduced an on-device text-to-image generation feature for Android using diffusion models, which has now been officially launched as the experimental Image Generator. This tool enables developers to generate images on Android devices in approximately 15 seconds, offering three main functionalities: text-to-image generation from prompts, controllable generation using text prompts and conditioning images via diffusion plugins, and customized generation with Low-Rank Adaptation (LoRA) weights for specific concepts. The Image Generator supports models compatible with Stable Diffusion v1.5 architecture and allows for fine-tuning through MediaPipe Diffusion LoRA on Vertex AI, offering developers the flexibility to customize foundation models with new concepts. The tool also includes a plugin system for enhanced generation capabilities with support for facial structures, edge detection, and depth awareness, as well as the ability to inject specific styles or objects into images using LoRA weights. Developers can explore these features through the official documentation and example app on GitHub, with ongoing support and updates available through Google’s platforms.