berkerbatur/speechready

Prepare text for text-to-speech. Cleans LLM output, verbalizes ambiguous tokens, compiles pauses and emphasis to each provider's syntax, and chunks under model