Models

• 🎙️ LKU & LKU v1.1 - Speech emotion recognition based on wav2vec2.

Data

• 😃 Kangurei 90K - 90k faces annotated with ARKit/PerfectSync blendshapes.
• 👻 OshiChats v2 - 56 million chat messages from VTuber live streams with smarter filtering, neural quality scores, and even more talents.
• 🎙️ LibriVox Tracks, a dataset of all 411K audio tracks uploaded to LibriVox before 26th September 2023, complete with reader ID & original text links.
• 👁️‍🗨️ OSHIChats v1 (August 2023), a dataset of 8 million high-quality chat messages collected and filtered from >1,000 VTuber live streams.