google-ai-edge/litert-samples

LiteRT and LiteRT-LM sample apps, model recipes, agent skills and utilities.

What it solves

This repository provides a comprehensive collection of sample applications, model recipes, and deployment tools for LiteRT (formerly TensorFlow Lite) and LiteRT-LM. It helps developers move machine learning models from training to on-device execution across Android, iOS, Python, and Web/WASM platforms, specifically focusing on maximizing hardware acceleration (GPU/NPU) and efficiency.

How it works

The project is organized into four main areas:

  • Application Samples: Runnable apps demonstrating various API paradigms, including the modern CompiledModel API for hardware acceleration, the legacy Interpreter API, and a specialized orchestration layer for LLMs (LiteRT-LM).
  • Model Recipes: Scripts and pipelines for converting and exporting models to the LiteRT format.
  • Utilities: Shared helper scripts and a GPU toolkit that optimizes PyTorch patterns for LiteRT GPU delegates.
  • Agent Skills: A set of automated "skills" that guide a model through the deployment lifecycle, covering conversion, quantization, on-device verification, and app scaffolding.

Who it’s for

Developers building on-device AI applications for mobile devices (Android/iOS) and web browsers who need practical examples of how to implement speech recognition, text-to-speech, image generation, and LLMs using Google's edge AI framework.

Highlights

  • Diverse Modalities: Includes samples for streaming TTS, ASR, multimodal vision-audio-text apps, and text-to-image diffusion.
  • LMM/LLM Support: Specialized samples for running Large Language Models via LiteRT-LM.
  • Hardware Optimization: Tools for GPU/NPU acceleration, including a GPU conversion toolkit and NPU JIT acceleration.
  • Deployment Lifecycle Automation: Agent-driven skills for accuracy-safe quantization and on-device verification.

Related

  • Project
    google-ai-edge/LiteRT

    Google's high-performance on-device runtime for deploying ML and Generative AI models across mobile, web, and IoT platforms with advanced hardware acceleration.

  • Project
    qualcomm/ai-hub-apps

    A collection of sample apps and tutorials for deploying machine learning models on Qualcomm devices, supporting various runtimes and Snapdragon NPU acceleration.

  • Project
    google-ai-edge/LiteRT-LM

    A production-ready orchestration layer for running high-performance LLMs on-device across mobile, web, and desktop platforms using LiteRT.

  • Project
    google-ai-edge/litert-torch

    A Python library for converting PyTorch models into .tflite format, enabling high-performance on-device execution on Android, iOS, and IoT devices.

  • Project
    Liquid4All/cookbook

    A comprehensive cookbook of examples, tutorials, and applications for deploying and fine-tuning Liquid AI's open-weight Liquid Foundation Models (LFMs) on edge devices, mobile, and browsers.