Return audio at the format it was given

app.py

@@ -355,7 +355,14 @@ def api_vamp(input_audio,
 
 @spaces.GPU
 def harp_vamp(input_audio, sampletemp, periodic_p, dropout, n_mask_codebooks, model_choice, stretch_factor):
-    sig = at.AudioSignal(input_audio).to_mono()
+    sig = at.AudioSignal(input_audio)
+
+    # Captured before the downmix, because generation runs in mono at VampNet's
+    # own rate and both have to be put back afterwards
+    original_sr = sig.sample_rate
+    original_channels = sig.num_channels
+
+    sig = sig.to_mono()
 
     input_audio = sig.cpu().detach().numpy()[0][0]
     input_audio = input_audio * np.iinfo(np.int16).max
@@ -386,9 +393,18 @@ def harp_vamp(input_audio, sampletemp, periodic_p, dropout, n_mask_codebooks, mo
         harp=True,
     )
 
+    sig = sig.detach().cpu()
+
+    if sig.sample_rate != original_sr:
+        sig = sig.resample(original_sr)
+
+    # Copies of one generated channel, not a stereo image, which the card says
+    if original_channels > 1 and sig.num_channels == 1:
+        sig.audio_data = sig.audio_data.repeat(1, original_channels, 1)
+
     ll = LabelList()
     ll.append(OutputLabel(label='short label', t=0.0, description='longer description'))
-    return save_audio(sig.detach().cpu()), ll
+    return save_audio(sig), ll
 
 
 if __name__ == "__main__":
@@ -797,14 +813,19 @@ if __name__ == "__main__":
             name="VampNet",
             description=(
                 (
-                    "Generates new audio by remixing and extending your input, filling in the "
-                    "blanks to create a coherent continuation in the same style. Available "
-                    "models include: cat10: fine-tuned on expressive cat sounds; "
+                    (
+                    "Generates new audio by remixing and extending your input, filling in "
+                    "the blanks to create a coherent continuation in the same style. "
+                    "Available models include: cat10: fine-tuned on expressive cat sounds; "
                     "lazaro-ros-sep: percussion extracted from the Lázaro Ros discography, "
-                    "rich in Afro-Cuban rhythm supermetroid: trained on Super Metroid (SNES) "
-                    "gameplay audio for retro, game-style soundscapes. Repository: "
+                    "rich in Afro-Cuban rhythm supermetroid: trained on Super Metroid "
+                    "(SNES) gameplay audio for retro, game-style soundscapes. Generation "
+                    "runs in mono at the model's own rate; the result is resampled back to "
+                    "the input's rate and copied across its channels, so the output has the "
+                    "format it came in with but no stereo image. Repository: "
                     "https://github.com/hugofloresgarcia/vampnet Paper: "
                     "https://arxiv.org/abs/2307.04686"
+                    )
                 )
             ),
             author="Hugo Flores Garcia, Prem Seetharaman, Rithesh Kumar, Bryan Pardo",