Upload folder using huggingface_hub

app.py

@@ -3,8 +3,6 @@ import os
 import numpy as np
 
 def no_cpu_warning(video_file, dataset, lora, size):
-    gr.Warning("This interface is running on CPU. Only cached examples are displayed here, no new music can be generated.")
-
     output = "./cached_examples/"
     output += "peft" if lora else "audiocraft"
     output += "_" + dataset + "_" + size + "_" + video_file[-7:-4]
@@ -13,38 +11,44 @@ def no_cpu_warning(video_file, dataset, lora, size):
 
     if not os.path.exists(output):
         print(output)
-        raise gr.Error("This combination of video and model has not been cached. Please use the other model or try one of the listed examples instead.")
+        raise gr.Error("This combination of video and model has not been cached. Try one of the listed examples below instead.")
 
+    gr.Warning("This interface is running on CPU. Only cached examples are displayed here, no novel music is generated. Please be patient, as the videos meight take a while to be downloaded.")
     print("Displaying video: " + output)
     return output
 
 interface = gr.Interface(fn=no_cpu_warning,
                          inputs=[
                                  gr.Video(value="videos/n_5.mp4",
-                                          label="Video Input",
+                                          label="Original Video",
                                           min_length=5,
                                           max_length=20,
-                                          sources=['upload'],
+                                          sources=[],
                                           show_download_button=True,
-                                          include_audio=True),
+                                          include_audio=True, 
+                                          mirror_webcam=False
+                                          ),
                                  gr.Radio(["nature", "symmv"],
                                           value="nature",
                                           label="Available Models",
-                                          info="Choose one of the available Datasets on which the models has been trained on."),
+                                          info="Choose one of the datasets on which the models has been trained on. Nature is a dataset of calm and relaxing sounds, symmv contains charts music. The model will generate audio that is similar to the training data."
+                                          ),
                                  gr.Radio([False, True],
                                           label="Use the LoRA version of the MusicGen Audio Decoder",
                                           value=False,
                                           info="If set to 'True' the MusicGen Audio Decoder models trained with LoRA "
                                                "(Low Rank Adaptation) are used. If set to 'False', the original "
-                                               "MusicGen models are used instead."),
+                                               "MusicGen models are used instead."
+                                               ),
                                  gr.Radio(["small", "medium", "large"],
                                           label="Model Size",
                                           value="large",
-                                          info="Choose one of the available model sizes. The larger models are more likely produce "
-                                               "results of higher audio quality, but also take more time to generate it."),
+                                          info="Choose one of the available model sizes. This effects the enocder and the deocder model. The larger models are more likely produce "
+                                               "results of higher audio quality, but also take more time to generate it."
+                                               ),
                                 ],
 
-                         outputs=[gr.Video(label="video output")],
+                         outputs=[gr.Video(label="Generated Result")],
                          examples=[
                              [os.path.abspath("./videos/n_1.mp4"), "nature", False, "large"],
                              [os.path.abspath("./videos/n_2.mp4"), "nature", False, "large"],
@@ -63,10 +67,11 @@ interface = gr.Interface(fn=no_cpu_warning,
                              [os.path.abspath("./videos/s_7.mp4"), "symmv", True, "large"],
                              [os.path.abspath("./videos/s_8.mp4"), "symmv", True, "large"],
                            ],
+                         examples_per_page=20,
                          cache_examples=False,
                          )
 
 if __name__ == "__main__":
     interface.launch(
-        share=True
+        share=False
     )