Blog
Teaching avatars to remember your face
We developed LoRAvatar, a new adaptation method to personalize avatars that preserve identity-specific details like wrinkles, tattoos, and subtle facial textures.
Atmanity Research
Teaching AI when to speak and how to react
We developed MM-When2Speak, a new multimodal AI model that uses video, audio, and text to deliver real-time conversational reactions, improving the naturalness of AI-human interaction through better timing.
Atmanity Research