Browser ML Demos

ONNX models running locally via Transformers.js and ONNX Runtime Web

Transformers.js Feature Explorer Patch similarity heatmap for a single image DINOv3 Transformers.js Cross-Image Correspondence Find matching patches between two images DINOv3 ONNX Runtime Depth Estimation Upload an image, get a colorized depth map Depth Anything v3 ONNX Runtime Neural Line Art Vectorization Convert raster line-art directly to clean vector SVG Line Art ONNX Runtime Portrait Relighting Change portrait lighting direction from a single image Relighting ONNX Runtime Colorify Colorize black & white photos and video Colorization ONNX Runtime FaceParser Semantic face segmentation powered by SegFormer Face Parsing ONNX Runtime ObjectDetect Real-time object detection in the browser Object Detection Transformers.js Vision Language Image understanding powered by Transformers.js Vision Language Transformers.js FastVLM Fast vision-language model with WebGPU acceleration Vision Language ONNX Runtime Revive Face restoration & colorization for degraded inputs Face Restoration ONNX Runtime InkTrace Convert any photo into clean contour line art Line Art ONNX Runtime Handwriting Synthesis Generate realistic cursive handwriting strokes from text input Handwriting ONNX Runtime PortraitCut Lightweight real-time portrait background removal & matting Background Removal Transformers.js PixelCaption Auto-generate natural language descriptions of any image with BLIP Image Captioning Transformers.js CLIPMatch Classify images using your own custom text labels — no fixed classes Zero-Shot CLIP Transformers.js OWLSearch Find any object by typing what to look for — open-vocabulary detection Text-Guided Detection Transformers.js PixelBoost AI-powered 2× image upscaling with Swin2SR super-resolution Super Resolution ONNX Runtime SwinIR Swin Transformer super-resolution — lightweight ×2 or classical ×4 Super Resolution ONNX Runtime StyleForge Apply artistic styles in real time Style Transfer Transformers.js SceneMap Semantic scene segmentation — every pixel labelled with SegFormer Segmentation Transformers.js SegClick Click anything in an image to instantly segment and extract it Segmentation Transformers.js Pocket Brain 360M-parameter LLM chat running fully in your browser Language Model Transformers.js VoiceScribe Real-time speech-to-text transcription powered by Moonshine ASR Audio Transformers.js TextSpeak Neural voice synthesis with 10 Kokoro voices — download WAV Audio ONNX Runtime TimbreShift Neural timbre transfer — change instrument timbre from any audio Audio

Coming Soon

Image Demoiréing Remove moiré patterns from screen photos up to 4K Restoration

Neural Painter Decompose photos into editable paintbrush stroke parameters Generative Art

Video Relighting Real-time 3D-aware portrait relighting at 33 FPS Relighting

Universal Relighting Relight any object using HDR environment maps Relighting

SVBRDF Estimation Estimate PBR material maps (albedo, normal, roughness) from one photo Relighting

Apple NeuralHash Perceptual hashing robust to resize, compression & edits Perceptual Hashing

Virtual Sketching Photo to vector strokes, sketch simplification & vectorization Line Art

Aesthetic Assessment Rate photo quality and aesthetic appeal with NIMA Assessment

Metric3D Depth Metric depth estimation with ViT backbone Depth / 3D

DETR End-to-end object detection with transformers Object Detection

YOLOv10 Real-time object detection, no NMS required Object Detection

Grounding DINO Open-set detection with text prompts Object Detection

BiRefNet High-resolution background removal & matting Segmentation

U²-Net Salient object & human/clothing segmentation Segmentation

ISNet Anime Anime character segmentation and extraction Segmentation

Swin2SR ×2 Classical image super-resolution at 2× Super Resolution

Swin2SR ×4 Real-world super-resolution at 4× with BSRGAN Super Resolution

DDColor Dual-decoder colorization for grayscale images Colorization

Style Transfer Apply artistic styles in real time Style Transfer

AnimeGANv2 Convert photos to anime-style artwork Style Transfer

DocShadow Remove shadows from scanned documents Document

ViT Classifier Image classification with Vision Transformer Classification

MobileViT Lightweight mobile vision transformer Classification

ResNet-50 Classic deep residual network classifier Classification

DistilBERT Sentiment Fast sentiment analysis on text NLP

BGE Reranker Cross-encoder reranking for search results NLP

Grammar Synthesis Grammar correction and text synthesis NLP

Chatterbox TTS Multilingual text-to-speech with ONNX Audio