Repolode
Submit
deepbeepmeep

deepbeepmeep/WanGP

A one-stop local app for open-source generative models, running video, image, audio, speech, 3D and interactive world generation on consumer GPUs.

Open-source alternative to:RunwayMidjourney

View on GitHub
Stars
11,437
Forks
1,854
Open issues
1,368
Last commit
20 hours ago

What is it?

WanGP is a one-stop app for open-source generative models. It bundles video models such as Wan 2.1/2.2, LTX-2, Hunyuan Video and MiniMax H3, image models such as Flux, Qwen Image and Krea 2, audio and TTS models, and several 3D reconstruction and generation models into a single interface. It runs locally on your own GPU, through a browser UI, or from the command line.

The project focuses on running large models on limited hardware. It offers memory profiles, quantized checkpoint formats (int8, fp8, GGUF, NV FP4, Nunchaku), and the README states it can run select models with as little as 4 GB of VRAM. A generation queue, headless batch mode, LoRA support, and an API and MCP interface let it fit into scripted or agent-driven workflows.

The bundled Deepy assistant can plan and run generation jobs in chat, and a Worlds gallery lets users walk through generated scenes with the keyboard. Media utilities cover mask editing, background removal, upscaling, frame interpolation and subtitle or transcription work.

Key features

  • 01Generates video, image, music, speech and 3D assets with many open-source model families.
  • 02Runs on consumer NVIDIA GPUs (GTX 10 series and newer) and supports AMD RDNA GPUs.
  • 03Includes memory profiles and quantized checkpoint formats to reduce VRAM use.
  • 04Provides a browser UI, a generation queue, headless batch mode, and an API and MCP interface.
  • 05Supports LoRA adapters and custom finetunes from Hugging Face or CivitAI.
  • 06Offers upscaling, frame interpolation, background removal and speech transcription tools.

Use cases

  • Generating short AI videos or images on a local GPU without cloud fees.
  • Running large open-source video models on a card with limited VRAM.
  • Batch-producing images, clips or audio from scripts or the command line.
  • Building 3D assets from a single photo or from text descriptions.
  • Exploring generated scenes interactively in the Worlds gallery.

Quick start

  1. 1

    Download the experimental WanGP Manager app for Windows from the releases page

  2. 2

    Use the guided setup in the Manager to pick a GPU profile and install Python and dependencies

Things to know

  • The one-click Manager is experimental and only covers Windows.
  • Local use needs a capable GPU; 8 GB of VRAM is recommended to get started.
  • The license could not be identified from the metadata, so check the repository's license terms before use.

In collections

AI video generationRun LLMs locallyAI video & audio toolsAI image generators & Stable Diffusion toolsRun and train AI on GPUs

Topics

#ai#video-generation#image-generation#text-to-speech#3d-generation#local-llm#gpu#diffusion-models#flux#flux2

FAQ

What is WanGP?
A one-stop local app for open-source generative models, running video, image, audio, speech, 3D and interactive world generation on consumer GPUs. WanGP is a one-stop app for open-source generative models. It bundles video models such as Wan 2.1/2.2, LTX-2, Hunyuan Video and MiniMax H3, image models such as Flux, Qwen Image and Krea 2, audio and TTS models, and several 3D reconstruction and generation models into a single interface. It runs locally on your own GPU, through a browser UI, or from the command line.
Is WanGP free to use? What license does it have?
WanGP is open source under the Other license. Check the license terms before using it in commercial products.
Is WanGP still maintained?
Actively maintained. The last commit to WanGP was on Nov 6, 2026.
What can I use WanGP for?
Generating short AI videos or images on a local GPU without cloud fees. Running large open-source video models on a card with limited VRAM. Batch-producing images, clips or audio from scripts or the command line. Building 3D assets from a single photo or from text descriptions. Exploring generated scenes interactively in the Worlds gallery.
How do I install WanGP?
Download the experimental WanGP Manager app for Windows from the releases page Use the guided setup in the Manager to pick a GPU profile and install Python and dependencies
Where can I get WanGP?
The source code and installation instructions are on GitHub: https://github.com/deepbeepmeep/WanGP

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects