web-txt2img

Generate images from text prompts directly in the browser using open-weights AI models. No server required - all inference runs locally using WebGPU acceleration.

Features

• 🚀 100% browser-based - No server, API keys, or network requests for inference
• 🎨 Multiple models - SD-Turbo (fast) and Janus-Pro-1B (quality)
• ⚡ WebGPU acceleration - Hardware-accelerated inference for fast generation
• 🔄 Worker-based - Non-blocking UI with progress tracking and cancellation
• 💾 Smart caching - Models cached locally after first download

Installation

npm i web-txt2img @xenova/transformers

For Janus-Pro-1B support, also install:

npm i @huggingface/transformers

Quick Start

import { Txt2ImgWorkerClient } from 'web-txt2img';

// Create worker client
const client = Txt2ImgWorkerClient.createDefault();

// Load SD-Turbo model with WebGPU
await client.load('sd-turbo', { 
  backendPreference: ['webgpu'] // WebGPU is required for reliable operation
});

// Generate image
const { promise } = client.generate({ 
  prompt: 'a cozy cabin in the woods, watercolor',
  seed: 42 
});

const result = await promise;
if (result.ok) {
  const url = URL.createObjectURL(result.blob);
  document.querySelector('img').src = url;
}

WebGPU Requirements

This library requires WebGPU support in your browser:

• Chrome/Edge 113+ with WebGPU enabled
• Safari Technology Preview with WebGPU feature flag
• Firefox Nightly with WebGPU enabled

You can verify WebGPU support:

const caps = await client.detect();
if (!caps.webgpu) {
  console.error('WebGPU not supported in this browser');
}

Supported Models

• sd-turbo - Fast single-step diffusion (512×512, ~2.3GB download)
  • Backend: WebGPU (required)
• janus-pro-1b - Higher quality autoregressive (~2.2GB)
  • Backend: WebGPU only

Documentation

• Full Documentation
• Live Demo
• Examples
• API Reference

Requirements

• Modern WebGPU-enabled browser (Chrome/Edge 113+, Safari Technology Preview, or Firefox Nightly with WebGPU)
• GPU with WebGPU support

Note on WASM Support

While a WASM fallback exists in the API for compatibility reasons, it is experimental, untested, and not recommended for production use. This library is designed and optimized for WebGPU.

License

MIT