Description

demos/musicgen_app.py

@@ -178,7 +178,7 @@ def predict_batched(texts, melodies):
     return res
 
 
-def predict_full(model, model_path, decoder, text, melody, duration, topk, topp, temperature, cfg_coef, progress=gr.Progress()):
+def predict_full(model, model_path, decoder, text, melody, duration, topk, topp, temperature, cfg_coef, output_hint, progress=gr.Progress()):
     global INTERRUPTING
     global USE_DIFFUSION
     INTERRUPTING = False
@@ -228,12 +228,19 @@ def predict_full(model, model_path, decoder, text, melody, duration, topk, topp,
 
 def toggle_audio_src(choice):
     if choice == "mic":
-        return gr.update(source="microphone", value=None, label="Microphone")
+        return gr.update(source = "microphone", value = None, label = "Microphone")
     else:
-        return gr.update(source="upload", value=None, label="File")
+        return gr.update(source = "upload", value = None, label = "File")
 
 
 def toggle_diffusion(choice):
+    if choice == "MultiBand_Diffusion":
+        return [gr.update(visible = True)] * 2
+    else:
+        return [gr.update(visible = False)] * 2
+
+
+def toggle_hint(choice):
     if choice == "MultiBand_Diffusion":
         return [gr.update(visible=True)] * 2
     else:
@@ -252,13 +259,18 @@ def ui_full(launch_kwargs):
             <p style="text-align: center;">Generates up to 2 minutes of music freely, without account and without watermark that you can download</p>
             <br/>
             <br/>
-            Derived from <a href="https://github.com/facebookresearch/audiocraft">MusicGen</a>,
+            Powered by <a href="https://github.com/facebookresearch/audiocraft">MusicGen</a>,
             presented at: <a href="https://huggingface.co/papers/2306.05284">"Simple and Controllable Music Generation"</a>.
             MusicGen gets better results than other AIs like WaveFormer.
             If you are looking for sound effect rather than music, I recommend you AudioGen or AudioLDM2.
             The generated tracks tend to be very monotone so I advise you to add an original track to force the AI to make variations.
+            <br/>
+            🐌 Slow process... ~6 hours for 2 minutes of music.
+            I advise you to use the identical <a href='https://huggingface.co/spaces/tomandandy/MusicGen5'>tomandandy's space</a> that runs faster on A10G.
             You can duplicate this space on a free account, it works on CPU.
             <a href="https://huggingface.co/spaces/Fabrice-TIERCELIN/Text-to-Music?duplicate=true"><img src="https://img.shields.io/badge/-Duplicate%20Space-blue?labelColor=white&style=flat&logo=data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAYAAAAf8/9hAAAAAXNSR0IArs4c6QAAAP5JREFUOE+lk7FqAkEURY+ltunEgFXS2sZGIbXfEPdLlnxJyDdYB62sbbUKpLbVNhyYFzbrrA74YJlh9r079973psed0cvUD4A+4HoCjsA85X0Dfn/RBLBgBDxnQPfAEJgBY+A9gALA4tcbamSzS4xq4FOQAJgCDwV2CPKV8tZAJcAjMMkUe1vX+U+SMhfAJEHasQIWmXNN3abzDwHUrgcRGmYcgKe0bxrblHEB4E/pndMazNpSZGcsZdBlYJcEL9Afo75molJyM2FxmPgmgPqlWNLGfwZGG6UiyEvLzHYDmoPkDDiNm9JR9uboiONcBXrpY1qmgs21x1QwyZcpvxt9NS09PlsPAAAAAElFTkSuQmCC&logoWidth=14"></a>
+            <br/>
+            🄯 If you use the standard models, you can use, modify and share the generated musics but not for commercial uses.
 
             """
         )
@@ -302,9 +314,10 @@ def ui_full(launch_kwargs):
                 diffusion_output = gr.Video(label="MultiBand Diffusion Decoder")
                 audio_diffusion = gr.Audio(label="MultiBand Diffusion Decoder (wav)", type='filepath')
         submit.click(toggle_diffusion, decoder, [diffusion_output, audio_diffusion], queue=False,
+                     show_progress=False).then(toggle_hint, decoder, [output_hint, audio_diffusion], queue=False,
                      show_progress=False).then(predict_full, inputs=[model, model_path, decoder, text, melody, duration, topk, topp,
-                                                                     temperature, cfg_coef],
-                                               outputs=[output, audio_output, diffusion_output, audio_diffusion])
+                                                                     temperature, cfg_coef, output_hint],
+                                               outputs=[output, audio_output, diffusion_output, audio_diffusion], scroll_to_output = True)
         radio.change(toggle_audio_src, radio, [melody], queue=False, show_progress=False)
 
         gr.Examples(