:beers: cheers

README.md

@@ -1,6 +1,6 @@
 ---
-title: Deepafx St
-emoji: 🐢
+title: Audio Style Transfer
+emoji: 🔥🗣️🎵🔥
 colorFrom: red
 colorTo: gray
 sdk: gradio

app.py

@@ -8,17 +8,24 @@ from huggingface_hub import hf_hub_download
 from deepafx_st.system import System
 from deepafx_st.utils import DSPMode
 
-system = System.load_from_checkpoint(
+system_speech = System.load_from_checkpoint(
     hf_hub_download("nateraw/deepafx-st-libritts-autodiff", "lit_model.ckpt"), batch_size=1
 ).eval()
+system_music = System.load_from_checkpoint(
+    hf_hub_download("nateraw/deepafx-st-jamendo-autodiff", "lit_model.ckpt"), batch_size=1
+).eval()
 
 gpu = torch.cuda.is_available()
 
 if gpu:
-    system.to("cuda")
+    system_speech.to("cuda")
+    system_music.to("cuda")
+
+
+def process(input_path, reference_path, model):
 
+    system = system_speech if model == "speech" else system_music
 
-def process(input_path, reference_path):
     # load audio data
     x, x_sr = torchaudio.load(input_path)
     r, r_sr = torchaudio.load(reference_path)
@@ -69,12 +76,18 @@ def process(input_path, reference_path):
 
 gr.Interface(
     fn=process,
-    inputs=[gr.Audio(type="filepath"), gr.Audio(type="filepath")],
+    inputs=[gr.Audio(type="filepath"), gr.Audio(type="filepath"), gr.Dropdown(["speech", "music"], value="speech")],
     outputs="audio",
     examples=[
         [
             hf_hub_download("nateraw/examples", "voice_raw.wav", repo_type="dataset", cache_dir="./data"),
             hf_hub_download("nateraw/examples", "voice_produced.wav", repo_type="dataset", cache_dir="./data"),
+            "speech",
+        ],
+        [
+            hf_hub_download("nateraw/examples", "nys_of_mind.wav", repo_type="dataset", cache_dir="./data"),
+            hf_hub_download("nateraw/examples", "world_is_yours_highpass.wav", repo_type="dataset", cache_dir="./data"),
+            "music",
         ],
     ],
     title="DeepAFx-ST",