WanGP

Web UI that runs Wan, Hunyuan, LTX, and Flux models on GPUs with as little as 6 GB of VRAM.

Self HostedOffline CapableGPU Required (6GB+ VRAM)
0.0 (0)

About

GPU-poor video generation is WanGP's entire mission: deepbeepmeep's web application, also known as Wan2GP, runs Wan 2.1 and 2.2, Hunyuan Video, LTX-Video, Flux, Qwen Image, and dozens of other open video, image, and audio models on cards with as little as 6 GB of VRAM, including RTX 10xx-era NVIDIA GPUs and AMD RDNA 2 through 4. It leans on aggressive memory management and quantized checkpoints in int8, fp8, GGUF, NVFP4, and Nunchaku formats, adds LoRA support, and wraps everything in a Gradio web UI with generation queues, galleries, built-in preprocessing such as masking and pose extraction, plus an API and MCP server for automation. Installation offers one-click Windows and Linux scripts, a Pinokio installer, a desktop launcher, Docker images, or manual conda setup. The software is free for personal, educational, and internal business use under the custom WanGP Community License 2.0, which forbids reselling it or offering it as a paid hosted service, making it source-available rather than strictly open source. Around 6,900 GitHub stars reflect its status as a default choice for local video generation on modest hardware.

Should you use WanGP?

Pick it when

Pick WanGP when you have a 6 to 12 GB NVIDIA card, even an RTX 10xx, or an AMD RDNA 2 to 4 GPU and want one web app running Wan, Hunyuan, LTX, and Flux with queues, LoRAs, and a one-click installer.

Look elsewhere when

Skip it if you plan to resell the software or offer it as a paid hosted service, which its license forbids, or if you need an OSI license. For custom node-graph pipelines, ComfyUI is the more flexible base.

Alternatives to WanGP

  • ComfyUI

    Node-based GPL-3.0 editor where video models such as Wan, LTX, and Kandinsky ship workflows, more flexible but with more wiring than WanGP's ready tabs.

  • SwarmUI

    MIT web UI on a ComfyUI backend with a simple Generate tab and multi-GPU server support, and its license does not bar hosted services.

  • FramePack

    Single-purpose Apache-2.0 app for long image-to-video on 6 GB, simpler if FramePack is all you need, without WanGP's multi-model menu.

  • Pinokio

    MIT one-click launcher that installs many AI apps, including WanGP itself, useful if you want several local tools managed in one place.

Reviews (0)

Leave a Review

No reviews yet. Be the first to review!

Details

Price
Free
Platform
Local/Desktop
Difficulty
Easy (2/5)
License
WanGP Community License 2.0
Minimum VRAM
6 GB
Added
Jul 29, 2026

Tags