Update app.py

app.py

@@ -381,8 +381,6 @@ class KittenTTSGradio:
             chunk_label = "chunk" if chunk_size == 1 else f"chunk ({chunk_size} sentences each)"
             progress(0, desc=f"Processing {total_sentences} sentences in {total_chunks} {chunk_label}s...")
             
-            audio_chunks = []
-            
             # Reset model state before starting
             if hasattr(self.model, 'session'):
                 try:
@@ -397,22 +395,26 @@ class KittenTTSGradio:
                 except:
                     pass
             
+            # Create a list to hold results in the correct order
+            audio_chunks = [None] * total_chunks
+            
             if use_multithreading and total_chunks > 1:
                 # Process chunks in parallel with limited workers
                 with ThreadPoolExecutor(max_workers=min(self.max_workers, 4)) as executor:
-                    futures = {
+                    # Submit all tasks
+                    future_to_index = {
                         executor.submit(self.process_single_sentence, chunk, voice, speed): i 
                         for i, chunk in enumerate(chunks)
                     }
                     
-                    results = {}
                     completed = 0
                     
-                    for future in as_completed(futures):
+                    # Process as they complete
+                    for future in as_completed(future_to_index):
+                        index = future_to_index[future]
                         try:
-                            idx = futures[future]
                             audio = future.result()
-                            results[idx] = audio
+                            audio_chunks[index] = audio  # Place at the correct index
                             completed += 1
                             progress(completed / total_chunks, 
                                    desc=f"Processed {completed}/{total_chunks} {chunk_label}s")
@@ -431,17 +433,21 @@ class KittenTTSGradio:
                                 except:
                                     pass
                         except Exception as e:
-                            print(f"Error processing chunk: {e}")
-                            continue
-                    
-                    for i in sorted(results.keys()):
-                        audio_chunks.append(results[i])
+                            print(f"Error processing chunk at index {index}: {e}")
+                            # Generate silence for failed chunks
+                            sample_rate = 24000
+                            silence_duration = 0.5
+                            silence = np.zeros(int(sample_rate * silence_duration))
+                            audio_chunks[index] = silence
+                            completed += 1
+                            progress(completed / total_chunks, 
+                                   desc=f"Processed {completed}/{total_chunks} {chunk_label}s")
             else:
                 # Process chunks sequentially
                 for i, chunk in enumerate(chunks):
                     try:
                         audio = self.process_single_sentence(chunk, voice, speed)
-                        audio_chunks.append(audio)
+                        audio_chunks[i] = audio
                         progress((i + 1) / total_chunks, 
                                desc=f"Processed {i + 1}/{total_chunks} {chunk_label}s")
                         
@@ -459,11 +465,25 @@ class KittenTTSGradio:
                             except:
                                 pass
                     except Exception as e:
-                        print(f"Error processing chunk: {e}")
-                        continue
+                        print(f"Error processing chunk at index {i}: {e}")
+                        # Generate silence for failed chunks
+                        sample_rate = 24000
+                        silence_duration = 0.5
+                        silence = np.zeros(int(sample_rate * silence_duration))
+                        audio_chunks[i] = silence
+                        progress((i + 1) / total_chunks, 
+                               desc=f"Processed {i + 1}/{total_chunks} {chunk_label}s")
             
-            if not audio_chunks:
-                raise gr.Error("Failed to generate any audio.")
+            # Check if we have any None values (shouldn't happen with the error handling)
+            if any(chunk is None for chunk in audio_chunks):
+                print("Warning: Some audio chunks were not generated properly")
+                # Replace any None values with silence
+                for i, chunk in enumerate(audio_chunks):
+                    if chunk is None:
+                        sample_rate = 24000
+                        silence_duration = 0.5
+                        silence = np.zeros(int(sample_rate * silence_duration))
+                        audio_chunks[i] = silence
             
             progress(0.9, desc="Concatenating audio...")