Velorn now finds models you keep in subfolders. If you organize your ComfyUI models into folders like models/diffusion_models/WAN/, everything just works now.
Fixed
Models organized into subfolders are now detected and used. Previously, a model living in a subfolder (say diffusion_models/WAN/wan2.2.safetensors) was reported as missing by the dependency checker, and generations could fail validation even when everything was installed - the first report came from a user whose tidy model library did exactly this. Velorn now matches models by filename anywhere ComfyUI can see them, automatically uses the correct subfolder path when queueing, and the on-disk check looks inside subfolders too (including folders wired up through extra_model_paths.yaml). If the same filename exists in several subfolders, Velorn picks deterministically and notes it in the generation log.
Thanks to the user who reported this within hours of trying Velorn - this is exactly the kind of feedback that makes the next person's first run smoother.
Notes
- Windows builds are code-signed as of v0.3.2 - if your antivirus flagged an older download, current releases install cleanly.
Downloads
Windows Installer: standard Windows install experience for most usersWindows Portable: no-install Windows build for quick testing or portable useMac (Apple Silicon): for M1, M2, M3, and newer MacsMac (Intel): for older Intel-based MacsLinux AppImage: portable Linux buildLinux deb: Debian/Ubuntu package
New here? See v0.3.0 for the audio mixer release and v0.2.0 for the MCP agent era.