Update app.py

app.py

@@ -7,7 +7,6 @@ import scipy
 from gtts import gTTS
 from io import BytesIO
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
-from transformers import VitsTokenizer, VitsModel
 
 
 def extract_text(pdf_file):
@@ -61,15 +60,11 @@ def text_to_audio(text):
     #buffer.seek(0)  
     #return buffer.read()
 
-    tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-eng")
-    model = VitsModel.from_pretrained("facebook/mms-tts-eng")
+    synthesiser = pipeline("text-to-speech", "suno/bark")
 
-    inputs = tokenizer([text], return_tensors="pt")
+    speech = synthesiser[str("summary"), forward_params={"do_sample": True}]
 
-    with torch.no_grad():
-      outputs = model(**inputs)
-  
-    return outputs.waveform[0]
+    scipy.io.wavfile.write("speech.wav", rate=speech["sampling_rate"], data=speech["audio"])
 
 def audio_pdf(pdf_file):
     text = extract_text(pdf_file)