subbed audio training data (much better than cc data) is better

		htrp 71 days ago \| parent \| context \| favorite \| on: Qwen3-TTS family is now open sourced: Voice design... subbed audio training data (much better than cc data) is better