Update app.py

app.py

@@ -1,20 +1,38 @@
 import gradio as gr
-from transformers import pipeline
+from transformers import pipeline, AutoProcessor, BarkModel
 import cv2
 from PIL import Image
 import io
 import scipy
+import torch
+
+
 
 #Commit
 def video_to_descriptions(video, target_language="en"):
+    
+    modelname="suno/bark-small"
+    processor = AutoProcessor.from_pretrained(hgmodelname)
+    model = BarkModel.from_pretrained(modelname, torch_dtype=torch.float16).to(device)
+    model = BetterTransformer.transform(model, keep_original_model=False)
+
+    # enable CPU offload
+    model.enable_cpu_offload()
+    
+    
+    
+    
+    
+
+    
     # Load the image-to-text and summarization pipelines
     ImgToText = pipeline("image-to-text", model="Salesforce/blip-image-captioning-large")
     Summarize = pipeline("summarization", model="tuner007/pegasus_summarizer")
     
     # Load the translation pipeline for the target language
     translator = pipeline("translation", model=f"Helsinki-NLP/opus-mt-en-{target_language}")
-    audio = pipeline("text-to-speech", model="suno/bark-small")
-    voice_preset = f"v2/{target_language}_speaker_2"
+    voice_preset = f"v2/{target_language}_speaker_1"
+    
     
     # Open the video
     cap = cv2.VideoCapture(video)
@@ -52,11 +70,16 @@ def video_to_descriptions(video, target_language="en"):
     translated_text = translator(summarized_description)[0]["translation_text"]
     print("TRANSLATION : " + translated_text)
     
-    audio_file = audio("translated_text", voice_preset=voice_preset)
-    print(audio_file)
+    inputs = processor(translated_text, voice_preset=voice_preset)
+    #audio_file = audio("translated_text", voice_preset=voice_preset)
+
+    
+    audio_array = model.generate(**inputs)
+    audio_array = audio_array.cpu().numpy().squeeze()
+    sample_rate = model.generation_config.sample_rate
 
     output_path = "./bark_out.wav"
-    scipy.io.wavfile.write(output_path, rate=audio_file["sampling_rate"], data=audio_file["audio"][0].squeeze())
+    scipy.io.wavfile.write(output_path, rate=sample_rate, data=audio_array)
     
     return output_path