---

title: Virtual Fitting Room emoji: 👕 colorFrom: red colorTo: pink sdk: gradio sdk_version: 5.34.2 app_file: app.py pinned: false

Virtual Fitting Room

This Hugging Face Space runs a CATVTON-based virtual fitting room demo. The person input supports both image upload and webcam capture. The clothing input uses image upload.

Required project structure

Upload these folders/files into the Space repository root together with app.py and requirements.txt:

• model/
• base_model/
• catvton_weights/
• utils.py

If your folders currently have long extracted names like:

• model-20260422T170440Z-3-001/model
• base_model-20260422T170410Z-3-002/base_model
• catvton_weights-20260422T170432Z-3-001/catvton_weights

rename or copy them in the Space repo root to exactly:

• model
• base_model
• catvton_weights

Notes

• Free CPU Spaces may be slow for high-quality inference.
• For better results, GPU hardware is recommended.
• If DensePose or SCHP dependencies are unavailable, the app falls back to a simpler mask.
• Webcam capture in the Space takes one still photo and then runs try-on. It is not frame-by-frame live video inference.

Connect the MVC app to this Space

After the Space is running, set your ASP.NET app configuration like this:

"VirtualTryOn": {
  "Mode": "HuggingFace",
  "HuggingFaceSpaceUrl": "https://YOUR_USERNAME-YOUR_SPACE_NAME.hf.space",
  "HuggingFaceApiName": "tryon",
  "HuggingFaceToken": ""
}

If the Space is private, create a Hugging Face access token and put it in the HF_TOKEN environment variable instead of committing it to appsettings.json.

The Gradio API endpoint is /call/tryon; the MVC service already sends the inputs in the order this Space expects.