Microsoft's VibeVoice text-to-speech model that can generate long-form speech from text with sample voices.
Want to make some of these yourself?