Update app.py

app.py

@@ -61,7 +61,7 @@ print(f"Using device: {device}, dtype: {dtype}")
 pipe = pipeline(
     "automatic-speech-recognition",
     model="openai/whisper-large-v3-turbo",
-    torch_dtype=torch.float32,
+    torch_dtype=dtype,
     device=device,
 )
 #vocos = Vocos.from_pretrained("charactr/vocos-mel-24khz")
@@ -344,7 +344,7 @@ def infer(ref_audio_orig, ref_text, gen_text, exp_name, remove_silence, cross_fa
     audio, sr = torchaudio.load(ref_audio)
 
     # Use the new chunk_text function to split gen_text
-    max_chars = int(len(ref_text.encode('utf-8')) / (audio.shape[-1] / sr) * (28 - audio.shape[-1] / sr))
+    max_chars = int(len(ref_text.encode('utf-8')) / (audio.shape[-1] / sr) * (25 - audio.shape[-1] / sr))
     gen_text_batches = chunk_text(gen_text, max_chars=135)
     print('ref_text', ref_text)
     for i, batch_text in enumerate(gen_text_batches):