quazim commited on
Commit
f503040
·
1 Parent(s): b2d3523
Files changed (1) hide show
  1. app.py +9 -14
app.py CHANGED
@@ -161,11 +161,12 @@ def generate_music(text_prompt, duration=10, guidance_scale=3.0):
161
  max_val = np.max(np.abs(audio_data))
162
  if max_val > 0:
163
  audio_data = audio_data / max_val * 0.95 # Scale to 95% to avoid clipping
164
-
165
- audio_data = audio_data.astype(np.float32)
166
-
167
  print(f"[GENERATION] Final audio shape: {audio_data.shape}")
168
- print(f"[GENERATION] Audio range: [{np.min(audio_data):.3f}, {np.max(audio_data):.3f}]")
 
169
  print(f"[GENERATION] Sample rate: {sample_rate}")
170
 
171
  timestamp = int(time.time() * 1000)
@@ -178,12 +179,13 @@ def generate_music(text_prompt, duration=10, guidance_scale=3.0):
178
  file_size = os.path.getsize(temp_path)
179
  print(f"[GENERATION] Audio saved to: {temp_path}")
180
  print(f"[GENERATION] File size: {file_size} bytes")
 
 
 
181
  else:
182
  print(f"[ERROR] Failed to create audio file: {temp_path}")
183
  return None
184
 
185
- return temp_path
186
-
187
  except Exception as e:
188
  print(f"[ERROR] Generation failed: {str(e)}")
189
  cleanup_gpu()
@@ -225,14 +227,7 @@ with gr.Blocks(title="MusicGen Large - Music Generation") as demo:
225
  with gr.Column():
226
  audio_output = gr.Audio(
227
  label="Generated Music",
228
- type="filepath",
229
- format="wav",
230
- interactive=False,
231
- autoplay=True,
232
- show_download_button=True,
233
- waveform_options=gr.WaveformOptions(
234
- show_recording_waveform=True
235
- )
236
  )
237
 
238
  with gr.Accordion("Tips", open=False):
 
161
  max_val = np.max(np.abs(audio_data))
162
  if max_val > 0:
163
  audio_data = audio_data / max_val * 0.95 # Scale to 95% to avoid clipping
164
+
165
+ audio_data = (audio_data * 32767).astype(np.int16). ###
166
+
167
  print(f"[GENERATION] Final audio shape: {audio_data.shape}")
168
+ print(f"[GENERATION] Audio range: [{np.min(audio_data)}, {np.max(audio_data)}]")
169
+ print(f"[GENERATION] Audio dtype: {audio_data.dtype}")
170
  print(f"[GENERATION] Sample rate: {sample_rate}")
171
 
172
  timestamp = int(time.time() * 1000)
 
179
  file_size = os.path.getsize(temp_path)
180
  print(f"[GENERATION] Audio saved to: {temp_path}")
181
  print(f"[GENERATION] File size: {file_size} bytes")
182
+
183
+ print(f"[GENERATION] Returning numpy tuple: ({sample_rate}, audio_array)")
184
+ return (sample_rate, audio_data)
185
  else:
186
  print(f"[ERROR] Failed to create audio file: {temp_path}")
187
  return None
188
 
 
 
189
  except Exception as e:
190
  print(f"[ERROR] Generation failed: {str(e)}")
191
  cleanup_gpu()
 
227
  with gr.Column():
228
  audio_output = gr.Audio(
229
  label="Generated Music",
230
+ type="numpy"
 
 
 
 
 
 
 
231
  )
232
 
233
  with gr.Accordion("Tips", open=False):