VibeVoice 1.5B Running Locally...On an iPhone! Only ~2.2 GB of Memory and Up to 1.28× Real-Time Speed

I speed up the generation part of the demo in case you get bored 😄

I also tested another long-form generation, and the VRAM usage looks stable. The demo is about a minute long, and I posted it on X.

This started as a random idea and somehow turned into a full detour from working on the next audio.cpp release. The model was uploaded to the audio.cpp HF repo. I will upload the xcframework later, and then push the code to a branch after release 0.6.