Ignore vocos for now

app.py

@@ -199,9 +199,9 @@ def generate(audio_path, ins, speed, alpha, beta, embedding, steps=100, vocos=Fa
         synthaud, s_prev = other_tts.long_inference_segment(i, diffusion_steps=steps,
                                                             alpha=alpha, beta=beta, is_phonemes=True,
                                                             embedding_scale=embedding, prev_s=s_prev, ref_s=ref_s,
-                                                            speed=speed, t=0.7, vocos=vocos)
+                                                            speed=speed, t=0.8)
 
-        n_trim = int(len(synthaud) * 0.008) # 960 samples
+        n_trim = int(len(synthaud) * 0.009)
         synthaud[:n_trim] = 0
         synthaud[-n_trim:] = 0
         audio = np.concatenate((audio, synthaud))