MooshieUI v2.0.1

One-click installer — just download and run. The app will automatically install ComfyUI, Python, PyTorch, and everything else you need on first launch.

Downloads

Platform	File
Windows	.msi or .exe
Linux	.deb or .AppImage

---

What's New in v2.0.1

New features

• Pick video frames straight from the gallery: "Make video" and "Add reference" now show up on the lightbox, the session hover bar, and the context menu for any still image, opening a gallery picker instead of requiring a file upload. A failed pick leaves whatever frame or reference was already set alone rather than clearing it.
• Model Stack shows the whole H3 tier: the Models view now lists every file a tier can use, both DiT variants included, with a per-file download row, instead of only the variant currently selected.

Fixes

• Video first/last frame distortion: a first or last frame image whose aspect ratio didn't match the output canvas is now pre-cropped to the canvas before generation, instead of being squashed or stretched to fit.
• Artist gallery: tags with no CDN preview are visible again: 7,415 artist tags that exist in the vocabulary but never got a CDN preview image (15 percent of the total) were invisible in the style explorer. They now show as placeholder cards behind an opt-in toggle, with a "generate this preview yourself" action that renders the same recipe the CDN previews use, locally.
• Server build: video_export.rs and video_interpolate.rs referenced tauri outside a desktop-feature gate, which broke the server-only binary build and skipped publishing it in v2.0.0's release CI. Fixed, and cargo check --no-default-features --features server is now a blocking gate in the pre-commit and release checks so it can't happen again silently.