deepbeepmeep/wan2gp ? reverse-engineered prompt

Build me a local web app like WanGP that lets me generate videos, images, and audio from one browser interface, with a clean queue so I can line up jobs and come back later. I want it to support a bunch of open source models for video, image, and text to speech, and to work well on modest GPUs, including older Nvidia cards and AMD where possible.

Please include a simple gallery to browse past generations, reuse outputs as new inputs, save reusable settings as templates, and let me manage model downloads and LoRAs without a lot of hassle. It should also have a few helpful prep tools for things like masks, background removal, pose or depth extraction, and basic post processing like upscaling and audio cleanup.

If you need to, look up current docs online for the latest model setup details, but keep the experience easy for non experts and make it feel fast and polished.