Update app.py

app.py

@@ -222,6 +222,8 @@ with gr.Blocks() as demo:
     Upload a skull-stripped FLAIR brain MRI in NIfTI (.nii.gz) format to generate a binary segmentation of multiple sclerosis lesions.  
     FLAMeS is based on the nnUNet framework<sup>2</sup> and was trained on 668 MRI scans acquired using Siemens, GE, and Philips 1.5T and 3T scanners<sup>1</sup>.
     For skull-stripping, we suggest using [SynthStrip](https://surfer.nmr.mgh.harvard.edu/docs/synthstrip/) with the `--no-csf` flag for optimal results.
+
+    Inference takes approximately 1 minute per MRI, with processing limited to one scan at a time due to Hugging Face's zero-GPU usage constraints. To process multiple cases simultaneously, download [FLAMeS's model](https://huggingface.co/FrancescoLR/FLAMeS-model) and run it locally using your own GPU or CPU setup.
     
     **Disclaimer:** Uploaded data is stored temporarily, no one has access to it, and it is deleted when the app is closed. For details, see [Gradio's file access guide](https://www.gradio.app/main/guides/file-access). Human subjects data should only be uploaded for processing if permitted by your institution's human subjects protection office.
     This is a research tool and is not intended for clinical use. Clinical decisions should not be based on the outputs of this tool.