Google AI Edge

LiteRT Sample Applications

Discover official sample applications, high-performance demos, and interactive environments designed for on-device machine learning acceleration.

All Library Samples

Developer Preview: These web demos are powered by the upcoming LiteRT Tensor API WASM library.

Web Playground
Live Web Demo

LiteRT Tensor API Sandbox

Run on-device machine learning execution directly in your browser. Test eager mode, static graphs, arithmetic reductions, and broadcasting using standard JavaScript console inputs.

Web Visualizer
Live Web Demo

Gemma 3 Tokenizer Demo

Interactive model visualizer exploring how Gemma 3 structures, tokenizes, and breaks down natural language text input in the browser using WebAssembly execution.

Web WebGPU
Live Web Demo

Image Segmentation (WebGPU)

Perform high-performance selfie multiclass image segmentation in real-time directly inside your web browser using hardware-accelerated WebGPU execution.

Web Simulation
Live Web Demo

Mandelbrot Fractal Generator

Highly optimized fractal canvas renderer demonstrating complex arithmetic operations executed using LiteRT Tensor WebAssembly backends.

Web Simulation
Live Web Demo

Conway's Game of Life

Classic cell automation board game implemented via 2D matrix computations triggered natively on-device with LiteRT WASM core functions.

Interpreter API

Image Classification

Identify and classify thousands of object categories inside images, live video streams, or custom user camera inputs. Supports standard mobile formats.

Interpreter API

Object Detection

Locate and draw bounding boxes around multiple distinct objects in a frame, complete with class name identification and trust percentage scores.

Interpreter API

Image Segmentation

Assign category labels to every pixel of an image, enabling background blurring, clean virtual backgrounds, and high-accuracy shape masking.

Interpreter API

Audio Classification

Record live audio samples using on-device microphones to identify sound events, environmental noises, music genres, and keyword voice commands.

Interpreter API

Digit Classifier (MNIST)

Recognize handwritten numerical digits drawn in real time onto canvas interfaces or parsed from paper document photography.

Interpreter API

Text Classification

Natural Language Processing (NLP) application analyzing sentence structures to classify input paragraphs, determine sentiments, and detect spam.

Interpreter API

Reinforcement Learning

Run on-device reinforcement learning agents to train game strategies, manage pathfindings, and control physics simulations in real-time.

Interpreter API

Super Resolution

Intelligently reconstruct low-resolution image assets to recover missing features, sharpen details, and produce high-definition visual outputs.

Compiled Model API

LLM Chatbot (Qualcomm NPU)

Accelerate large language model text generation natively on Qualcomm NPU hardware. Leverage async execution pipelines for fast on-device GenAI conversational response times.

Interactive Jupyter

LiteRT AOT Compilation

Hands-on tutorial demonstrating Ahead-of-Time (AOT) compilation strategies to target specialized GPU/NPU acceleration modules, boosting inference execution speed.

Compiled Model API

Accelerated Speech Recognition

Automatic Speech Recognition (ASR) converting continuous audio input streams into transcribed text paragraphs using hardware-compiled models.

Compiled Model API

Semantic Similarity API

Calculate mathematical distances and similarity metrics between text embeddings instantly, optimized for advanced client search pipelines.

Compiled Model API

Accelerated Text Classification

Process real-time texts, spam detections, and mood-metric classification models using CompiledModel API backends.

Compiled Model API

Accelerated Image Classification

Demonstrate state-of-the-art high frequency GPU/NPU image classification executing with async queues and optimized buffer allocations.

Compiled Model API

Accelerated Image Segmentation

High-throughput semantic segmentation of images. Enables rich camera masks and background isolation with specialized accelerators.

End-to-End Sample

ImageNet Classification

Comprehensive full system implementation showing end-to-end image parsing, preprocessing, and classification using standard ImageNet benchmarks.

No samples found

Try refining your search terms or choosing a different category tab.