11

tts_model.py

@@ -79,7 +79,7 @@ def generate_audio_for_seed(chat, seed, texts, batch_size, speed, refine_text_pr
             raise ValueError("维度应为 768。")
     else:
         raise ValueError(f"Invalid speaker_type: {speaker_type}. ")
-
+    print(f'generate_audio_for_seed,1, speaker_type:{speaker_type}')
     params_infer_code = {
         'spk_emb': rnd_spk_emb,
         'prompt': f'[speed_{speed}]',
@@ -99,7 +99,7 @@ def generate_audio_for_seed(chat, seed, texts, batch_size, speed, refine_text_pr
     flag = 0
     if not cur_tqdm:
         cur_tqdm = tqdm
-
+    print(f'generate_audio_for_seed,2, speaker_type:{speaker_type}')
     if re.search(r'\[uv_break\]|\[laugh\]', ''.join(texts)) is not None:
         if not skip_refine_text:
             print("Detected [uv_break] or [laugh] in text, skipping refine_text")
@@ -112,6 +112,7 @@ def generate_audio_for_seed(chat, seed, texts, batch_size, speed, refine_text_pr
                           use_decoder=True, skip_refine_text=skip_refine_text)
         all_wavs.extend(wavs)
         clear_cuda_cache()
+    print(f'generate_audio_for_seed,3, speaker_type:{speaker_type}')
     if skip_save:
         return all_wavs
     combined_audio = combine_audio(all_wavs)