Update app.py

app.py

@@ -337,7 +337,7 @@ chain_neo4j = (
 
 
 
-
+#perfect
 
 
 def bot(history, choice, tts_choice, retrieval_mode, model_choice):
@@ -369,6 +369,8 @@ def bot(history, choice, tts_choice, retrieval_mode, model_choice):
     history.append([response, None])
 
 
+#Bot response withstream and flicker with time 
+
 # def bot(history, choice, tts_choice, retrieval_mode, model_choice):
 #     if not history:
 #         return history
@@ -771,6 +773,8 @@ def generate_audio_elevenlabs(text):
         return None
 
 
+#Normal cases 
+
 # from parler_tts import ParlerTTSForConditionalGeneration, ParlerTTSStreamer
 # from transformers import AutoTokenizer
 # from threading import Thread
@@ -836,78 +840,7 @@ def generate_audio_elevenlabs(text):
 #     return combined_audio_path
 
 
-
-# import concurrent.futures
-# import tempfile
-# import os
-# import numpy as np
-# import logging
-# from queue import Queue
-# from threading import Thread
-# from scipy.io.wavfile import write as write_wav
-# from parler_tts import ParlerTTSForConditionalGeneration, ParlerTTSStreamer
-# from transformers import AutoTokenizer
-
-# # Ensure your device is set to CUDA
-# device = "cuda:0" if torch.cuda.is_available() else "cpu"
-
-# repo_id = "parler-tts/parler-tts-mini-v1"
-
-# def generate_audio_parler_tts(text):
-#     description = "A female speaker delivers a slightly expressive and animated speech with a moderate speed and pitch. The recording is of very high quality, with the speaker's voice sounding clear and very close up."
-#     chunk_size_in_s = 0.5
-
-#     # Initialize the tokenizer and model
-#     parler_tokenizer = AutoTokenizer.from_pretrained(repo_id)
-#     parler_model = ParlerTTSForConditionalGeneration.from_pretrained(repo_id).to(device)
-#     sampling_rate = parler_model.audio_encoder.config.sampling_rate
-#     frame_rate = parler_model.audio_encoder.config.frame_rate
-
-#     def generate(text, description, play_steps_in_s=0.5):
-#         play_steps = int(frame_rate * play_steps_in_s)
-#         streamer = ParlerTTSStreamer(parler_model, device=device, play_steps=play_steps)
-
-#         inputs = parler_tokenizer(description, return_tensors="pt").to(device)
-#         prompt = parler_tokenizer(text, return_tensors="pt").to(device)
-
-#         generation_kwargs = dict(
-#             input_ids=inputs.input_ids,
-#             prompt_input_ids=prompt.input_ids,
-#             attention_mask=inputs.attention_mask,
-#             prompt_attention_mask=prompt.attention_mask,
-#             streamer=streamer,
-#             do_sample=True,
-#             temperature=1.0,
-#             min_new_tokens=10,
-#         )
-
-#         thread = Thread(target=parler_model.generate, kwargs=generation_kwargs)
-#         thread.start()
-
-#         for new_audio in streamer:
-#             if new_audio.shape[0] == 0:
-#                 break
-#             # Save or process each audio chunk as it is generated
-#             yield sampling_rate, new_audio
-
-#     audio_segments = []
-#     for (sampling_rate, audio_chunk) in generate(text, description, chunk_size_in_s):
-#         audio_segments.append(audio_chunk)
-            
-#         temp_audio_path = os.path.join(tempfile.gettempdir(), f"parler_tts_audio_chunk_{len(audio_segments)}.wav")
-#         write_wav(temp_audio_path, sampling_rate, audio_chunk.astype(np.float32))
-#         logging.debug(f"Saved chunk to {temp_audio_path}")
-        
-
-#     # Combine all the audio chunks into one audio file
-#     combined_audio = np.concatenate(audio_segments)
-#     combined_audio_path = os.path.join(tempfile.gettempdir(), "parler_tts_combined_audio_stream.wav")
-
-#     write_wav(combined_audio_path, sampling_rate, combined_audio.astype(np.float32))
-
-#     logging.debug(f"Combined audio saved to {combined_audio_path}")
-#     return combined_audio_path
-
+# chunking audio and then Process
 
 import concurrent.futures
 import tempfile
@@ -917,8 +850,8 @@ import logging
 from queue import Queue
 from threading import Thread
 from scipy.io.wavfile import write as write_wav
-from transformers import AutoTokenizer
 from parler_tts import ParlerTTSForConditionalGeneration, ParlerTTSStreamer
+from transformers import AutoTokenizer
 
 # Ensure your device is set to CUDA
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
@@ -959,22 +892,97 @@ def generate_audio_parler_tts(text):
         for new_audio in streamer:
             if new_audio.shape[0] == 0:
                 break
-            # Save each audio chunk as it is generated
-            temp_audio_path = os.path.join(tempfile.gettempdir(), f"parler_tts_audio_chunk_{len(audio_segments)}.wav")
-            write_wav(temp_audio_path, sampling_rate, new_audio.astype(np.float32))
-            yield temp_audio_path
+            # Save or process each audio chunk as it is generated
+            yield sampling_rate, new_audio
 
     audio_segments = []
-    for audio_chunk_path in generate(text, description, chunk_size_in_s):
-        audio_segments.append(audio_chunk_path)
-        yield audio_chunk_path  # Yield each audio chunk path immediately
+    for (sampling_rate, audio_chunk) in generate(text, description, chunk_size_in_s):
+        audio_segments.append(audio_chunk)
+            
+        temp_audio_path = os.path.join(tempfile.gettempdir(), f"parler_tts_audio_chunk_{len(audio_segments)}.wav")
+        write_wav(temp_audio_path, sampling_rate, audio_chunk.astype(np.float32))
+        logging.debug(f"Saved chunk to {temp_audio_path}")
+        
 
     # Combine all the audio chunks into one audio file
     combined_audio = np.concatenate(audio_segments)
     combined_audio_path = os.path.join(tempfile.gettempdir(), "parler_tts_combined_audio_stream.wav")
+
     write_wav(combined_audio_path, sampling_rate, combined_audio.astype(np.float32))
 
     logging.debug(f"Combined audio saved to {combined_audio_path}")
+    return combined_audio_path
+
+
+# #streaming code in chunk
+
+
+# import concurrent.futures
+# import tempfile
+# import os
+# import numpy as np
+# import logging
+# from queue import Queue
+# from threading import Thread
+# from scipy.io.wavfile import write as write_wav
+# from transformers import AutoTokenizer
+# from parler_tts import ParlerTTSForConditionalGeneration, ParlerTTSStreamer
+
+# # Ensure your device is set to CUDA
+# device = "cuda:0" if torch.cuda.is_available() else "cpu"
+
+# repo_id = "parler-tts/parler-tts-mini-v1"
+
+# def generate_audio_parler_tts(text):
+#     description = "A female speaker delivers a slightly expressive and animated speech with a moderate speed and pitch. The recording is of very high quality, with the speaker's voice sounding clear and very close up."
+#     chunk_size_in_s = 0.5
+
+#     # Initialize the tokenizer and model
+#     parler_tokenizer = AutoTokenizer.from_pretrained(repo_id)
+#     parler_model = ParlerTTSForConditionalGeneration.from_pretrained(repo_id).to(device)
+#     sampling_rate = parler_model.audio_encoder.config.sampling_rate
+#     frame_rate = parler_model.audio_encoder.config.frame_rate
+
+#     def generate(text, description, play_steps_in_s=0.5):
+#         play_steps = int(frame_rate * play_steps_in_s)
+#         streamer = ParlerTTSStreamer(parler_model, device=device, play_steps=play_steps)
+
+#         inputs = parler_tokenizer(description, return_tensors="pt").to(device)
+#         prompt = parler_tokenizer(text, return_tensors="pt").to(device)
+
+#         generation_kwargs = dict(
+#             input_ids=inputs.input_ids,
+#             prompt_input_ids=prompt.input_ids,
+#             attention_mask=inputs.attention_mask,
+#             prompt_attention_mask=prompt.attention_mask,
+#             streamer=streamer,
+#             do_sample=True,
+#             temperature=1.0,
+#             min_new_tokens=10,
+#         )
+
+#         thread = Thread(target=parler_model.generate, kwargs=generation_kwargs)
+#         thread.start()
+
+#         for new_audio in streamer:
+#             if new_audio.shape[0] == 0:
+#                 break
+#             # Save each audio chunk as it is generated
+#             temp_audio_path = os.path.join(tempfile.gettempdir(), f"parler_tts_audio_chunk_{len(audio_segments)}.wav")
+#             write_wav(temp_audio_path, sampling_rate, new_audio.astype(np.float32))
+#             yield temp_audio_path
+
+#     audio_segments = []
+#     for audio_chunk_path in generate(text, description, chunk_size_in_s):
+#         audio_segments.append(audio_chunk_path)
+#         yield audio_chunk_path  # Yield each audio chunk path immediately
+
+#     # Combine all the audio chunks into one audio file
+#     combined_audio = np.concatenate(audio_segments)
+#     combined_audio_path = os.path.join(tempfile.gettempdir(), "parler_tts_combined_audio_stream.wav")
+#     write_wav(combined_audio_path, sampling_rate, combined_audio.astype(np.float32))
+
+#     logging.debug(f"Combined audio saved to {combined_audio_path}")