Note for anyone trying my Spaces:
The apps are configured to run on CPU so they work within Hugging Face's free tier limits. If you want faster performance, download the app.py file, install it locally, and swap tensorflow-cpu โ tensorflow (or add device=0 instead of device=-1 in the pipeline calls for torch-based apps). A machine with even a modest GPU will run inference 5โ10x faster. All the code is commented so the swap is a one-line change.