Update app.py

app.py

@@ -73,7 +73,7 @@ vocos = load_vocoder()
 
 target_sample_rate = 24000
 n_mel_channels = 100
-hop_length = 256
+hop_length = 256 - 8
 target_rms = 0.1
 nfe_step = 16  # 16, 32
 cfg_strength = 2.0
@@ -243,7 +243,7 @@ def infer_batch(ref_audio, ref_text, gen_text_batches, exp_name, remove_silence,
         final_text_list = convert_char_to_pinyin(text_list)
 
         # Calculate reference audio length
-        ref_audio_len = audio.shape[-1] + 8 // hop_length
+        ref_audio_len = audio.shape[-1] // hop_length
 
         if fix_duration is not None:
             duration = int(fix_duration * target_sample_rate / hop_length)