How is inference performed using a cascaded approach to generate speech from text?

[talkking]

Ming/test_audio_tasks.py

Lines 159 to 161 in 480df09

	for tts_speech, text_list in model.talker.omni_audio_generation(
	output_text, audio_detokenizer=audio_detokenizer, thinker_reply_part=thinker_reply_part, speaker=speaker, stream=stream, **spk_input
	):

Are you using a cascaded approach to generate the speech?