Everything in the catalog

Dia - 1.6B Text-to-Dialogue Model

Runs Chandra, Nanonets, olmOCR and Dots OCR models on a page image to extract text.

Lumian-VLR / VisionThink / MiniCPM-V / Typhoon-OCR / olmOCR

Converts a floor plan image into a digital layout with detected walls and rooms.

Lightweight OCR demo based on ONNX models that reads text from images.

A Family of Open Sourced Music Foundation Models

8M-param English TTS distilled from Kokoro-82M

Classic Tesseract OCR: upload an image and get the recognized text.

Generates subtitle files from a video or audio file using Whisper.

Converts an uploaded image into a pencil-style drawing.

Compare coding-agent harnesses, every claim linked to source

Video generation with a synchronized soundtrack

Ask questions about images and videos with the 32B Qwen2.5-VL model.

Generate structured output such as JSON from an in-browser WebLLM model.

Reads Japanese text from an image, including manga-style text.

Checks English sentences for errors and rewrites the problem ones with a FLAN-T5 model.

Moves and resizes an object inside an image without a LoRA, using a turbo Qwen-Image-2.1 setup.

Twelve camera, drawing and text demos of d1-2B

Creates songs, arranges vocals and remixes music using StepFun models.

Turns a single photo into a video where the camera appears to orbit 360 degrees around a frozen moment.

Minimal chat page where a Qwen2 model runs on your GPU in the browser using Transformers.js.

Translates text between languages with a model that runs in the browser.

Shows a live depth map of your webcam feed computed on the GPU in the browser.

Removes the background from webcam video frames in real time using a WebGPU model.