Skip to content

LightX2V Documentation

Lighter, faster image and video inference.

Turn text and images into visual content. LightX2V brings model execution, inference acceleration and deployment into a lightweight framework, from your first generation to a working application.

Get started Choose a model

Generate images and videos

Find a model for text-to-image, image editing, video or audiovisual generation.

Browse models →

Combine acceleration methods

Tune inference with quantization, feature caching, attention optimization and step distillation.

Choose an acceleration approach →

Work with available memory

Use parameter offload, quantization and parallel configurations to fit your hardware.

See low-resource deployment →

Connect apps and services

Build generation workflows with the Python API, inference services, Gradio or ComfyUI.

Set up an inference service →

Start here

Choose an entry point for your first generation.

Generate your first video

Prepare the environment, download Wan2.1-T2V-1.3B, and generate a video with the provided script.

  1. 1Set up
  2. 2Get weights
  3. 3Generate

Open the quickstart

Use the Python API
Create a pipeline and reuse it in your application.

Choose a model by task
Find its weights, scripts and configurations.

Use a graphical interface
Set up an interactive generation UI with Gradio.

Explore models by task

Start with what you want to create, then find the models and examples for that task.

Video and audioWan, MiniMax-H3, LTX and more Image generation and editingQwen-Image, Z-Image, FLUX.2 and more Restoration and upscalingSeedVR2, SwiftVR, RelayVSR Interaction and spatial tasksEdenLive, WorldMirror, WorldPlay

View the model catalog

Recent updates

New autoregressive video model entry points
Scripts and configurations for TaoMate-H3, StreamRender-H3, Matrix-Game-4 and more.

Interactive video generation with EdenLive
Control generation with text, a target first frame and proxy recordings.

Run LightX2V on a MacBook
Generate and edit images locally with a native macOS app.

Help improve LightX2V

Documentation issues · Inference issues · Get model weights · Read the technical blog