Update app.py

app.py

@@ -506,6 +506,32 @@ def generate_audio_elevenlabs(text):
         logging.error(f"Error generating audio: {response.text}")
         return None
 
+# def generate_audio_parler_tts(text):
+#     model_id = 'parler-tts/parler_tts_mini_v0.1'
+#     device = "cuda:0" if torch.cuda.is_available() else "cpu"
+#     try:
+#         model = ParlerTTSForConditionalGeneration.from_pretrained(model_id).to(device)
+#     except torch.cuda.OutOfMemoryError:
+#         print("CUDA out of memory. Switching to CPU.")
+#         device = "cpu"
+#         model = ParlerTTSForConditionalGeneration.from_pretrained(model_id).to(device)
+#     tokenizer = AutoTokenizer.from_pretrained(model_id)
+
+#     description = "A female speaker with a slightly low-pitched voice delivers her words quite expressively, in a very confined sounding environment with clear audio quality. She speaks very fast."
+
+#     input_ids = tokenizer(description, return_tensors="pt").input_ids.to(device)
+#     prompt_input_ids = tokenizer(text, return_tensors="pt").input_ids.to(device)
+
+#     generation = model.generate(input_ids=input_ids, prompt_input_ids=prompt_input_ids)
+#     audio_arr = generation.cpu().numpy().squeeze()
+    
+#     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as f:
+#         sf.write(f.name, audio_arr, model.config.sampling_rate)
+#         temp_audio_path = f.name
+
+#     logging.debug(f"Audio saved to {temp_audio_path}")
+#     return temp_audio_path
+
 def generate_audio_parler_tts(text):
     model_id = 'parler-tts/parler_tts_mini_v0.1'
     device = "cuda:0" if torch.cuda.is_available() else "cpu"
@@ -520,18 +546,31 @@ def generate_audio_parler_tts(text):
     description = "A female speaker with a slightly low-pitched voice delivers her words quite expressively, in a very confined sounding environment with clear audio quality. She speaks very fast."
 
     input_ids = tokenizer(description, return_tensors="pt").input_ids.to(device)
-    prompt_input_ids = tokenizer(text, return_tensors="pt").input_ids.to(device)
+    max_length = model.config.max_length
 
-    generation = model.generate(input_ids=input_ids, prompt_input_ids=prompt_input_ids)
-    audio_arr = generation.cpu().numpy().squeeze()
+    # Split the text into smaller chunks if it exceeds the max length
+    text_chunks = [text[i:i+max_length] for i in range(0, len(text), max_length)]
+    audio_segments = []
+
+    for chunk in text_chunks:
+        prompt_input_ids = tokenizer(chunk, return_tensors="pt").input_ids.to(device)
+        generation = model.generate(input_ids=input_ids, prompt_input_ids=prompt_input_ids)
+        audio_arr = generation.cpu().numpy().squeeze()
+        audio_segments.append(audio_arr)
     
+    combined_audio = np.concatenate(audio_segments)
+
     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as f:
-        sf.write(f.name, audio_arr, model.config.sampling_rate)
+        sf.write(f.name, combined_audio, model.config.sampling_rate)
         temp_audio_path = f.name
 
     logging.debug(f"Audio saved to {temp_audio_path}")
     return temp_audio_path
 
+
+
+
+
 # Stable Diffusion setup
 pipe = StableDiffusion3Pipeline.from_pretrained("stabilityai/stable-diffusion-3-medium-diffusers", torch_dtype=torch.float16)
 pipe = pipe.to("cuda")