updating result output to audio file rather than encoded wav

demos/melodyflow_app.py

@@ -20,7 +20,6 @@ from tempfile import NamedTemporaryFile
 import time
 import typing as tp
 import warnings
-import base64
 import asyncio
 import threading
 from concurrent.futures import ThreadPoolExecutor
@@ -281,16 +280,7 @@ def _do_predictions_batch(texts, melodies, solver, steps, target_flowstep,
                     file.name, output, MODEL.sample_rate, strategy="loudness",
                     loudness_headroom_db=16, loudness_compressor=True, add_suffix=False)
                 
-                # Read and encode audio
-                with open(file.name, 'rb') as f:
-                    audio_bytes = f.read()
-                audio_base64 = base64.b64encode(audio_bytes).decode('utf-8')
-                
-                results.append({
-                    "audio": audio_base64,
-                    "format": "wav"
-                })
-                
+                results.append(file.name)
                 file_cleaner.add(file.name)
         
         print(f"Batch finished: {len(texts)} requests in {time.time() - be:.2f}s")
@@ -461,6 +451,8 @@ def predict(model, text,
     # Get result
     try:
         result = future.result()
+        if isinstance(result, list) and len(result) > 0:
+            return result[0]
         return result
     except Exception as e:
         raise gr.Error(f"Generation failed: {str(e)}")
@@ -530,7 +522,7 @@ def ui_local(launch_kwargs):
                         label="Regularization Strength", minimum=0.0, maximum=1.0, value=0.0, interactive=False)
             with gr.Column():
                 audio_outputs = [
-                    gr.JSON(label=f"Generated Audio - variation {i+1}") for i in range(N_REPEATS)]
+                    gr.Audio(label=f"Generated Audio - variation {i+1}") for i in range(N_REPEATS)]
         submit.click(fn=predict,
                      inputs=[model, text,
                              solver,
@@ -541,7 +533,7 @@ def ui_local(launch_kwargs):
                              duration,
                              melody,
                              model_path,],
-                     outputs=[o for o in audio_outputs],
+                     outputs=audio_outputs,
                      concurrency_limit=8)  # Set concurrency limit on the event listener
         melody.change(toggle_melody, melody, [solver])
         solver.change(toggle_solver, [solver, melody], [steps, target_flowstep,
@@ -653,7 +645,7 @@ def ui_hf(launch_kwargs):
                         label="Regularization Strength", minimum=0.0, maximum=1.0, value=0.0, interactive=False)
             with gr.Column():
                 audio_outputs = [
-                    gr.JSON(label=f"Generated Audio - variation {i+1}") for i in range(N_REPEATS)]
+                    gr.Audio(label=f"Generated Audio - variation {i+1}") for i in range(N_REPEATS)]
         submit.click(fn=predict,
                      inputs=[model, text,
                              solver,
@@ -663,7 +655,7 @@ def ui_hf(launch_kwargs):
                              regularization_strength,
                              duration,
                              melody,],
-                     outputs=[o for o in audio_outputs],
+                     outputs=audio_outputs,
                      concurrency_limit=8)  # Set concurrency limit on the event listener
         melody.change(toggle_melody, melody, [solver])
         solver.change(toggle_solver, [solver, melody], [steps, target_flowstep,