I just tried VibeVoice on my local machine, and it’s pretty interesting. I fed three 15 second audio clips with three different voices which you can hear little snippets in the beginning, along with a podcast transcript generated from ChatGPT, and got this AI slop. It’s just a one shot run with no editing. The quality isn’t quite at NotebookLM’s level, but it’s still pretty cool to see open-source TTS can achieve this locally. #TTS #VibeVoice @ZBennoui github.com/microsoft/VibeVoice