Website Home (ChinaZ.com) June 14 News: Researchers from Stanford University, in collaboration with Apparate Labs, have introduced an AI model named PROTEUS. This model can generate realistic, expressive virtual characters in real-time from a single photo, enabling them to sing and speak.
Key Features:
Real-time generation of realistic characters: PROTEUS can generate effects such as laughing, rapping, singing, blinking, smiling, and speaking from a single image, showcasing complex facial expressions and body movements.
High frame rate video stream: Supports video streams of over 100 FPS, enabling real-time processing and ensuring smooth and natural interactions.
Multimodal interaction: Compatible with various data forms including voice, text, and images, allowing for natural and intuitive interactions in different scenarios.
Customization and application: Highly customizable architecture designed for multiple fields and application scenarios, meeting personalized needs.
Technical Architecture:
PROTEUS employs a latent diffusion model and advanced Transformer architecture, efficiently generating complex images by processing data in latent space.
Further improvements in architecture and algorithms have achieved generation speeds of over 100 frames per second.
Application Scenarios:
Personalized virtual assistants: Offering services such as daily task management, scheduling, and information inquiries.
Virtual pets: Creating virtual pets with realistic appearances and rich emotions.
Emotional support: Generating emotionally supportive virtual characters for psychological comfort and support.
Customer service: Creating virtual customer service representatives for instant and efficient customer support.
Education and training: Generating virtual teachers or trainers for personalized education and training.
Video game character customization: Providing highly customizable game characters for developers.
Film and entertainment: Used for generating realistic virtual actors and characters, reducing production costs.
Marketing and advertising: Generating virtual spokespersons for product promotion and brand advocacy.
Social media and virtual social interactions: Generating virtual avatars on social platforms to enrich social experiences.
PROTEUS aims to provide a visual representation that can be controlled by voice, serving as an intuitive interface for artificial conversational entities, allowing users to engage in natural dialogue and interaction with virtual characters. The security of this technology and early API access will be available to selected developers.
PROTEUS has been showcased in multiple applications on Twitch, demonstrating its potential in real-time interactive scenarios. Through the API, PROTEUS can be invoked in any application, bringing innovative virtual character interaction experiences to various industries.
Official Website: https://apparate.ai/stream.html
