3D digital human behavior AI engine
- An efficiently built highly intelligent and expressive 3D digital human behavior-driven AI engine. Combined with deep learning technology, it uses the Transformer architecture's self-attention mechanism to realize seamless transition from speech and text to high-quality mouth animation.
- Powerful animation reasoning ability and real-time rendering engine are deeply combined, using MHC Talker Unreal Engine plug-in to give the character "human-like" dialog interaction ability.
- From natural language understanding, emotional expression interpretation, to body movement logic response, all can be quickly configured to easily and efficiently realize high-quality character dialogue interaction, providing efficient immersive 3D digital character interaction solutions for digital human AI Agent, game intelligent NPC, virtual idol IP, meta-universe socialization, film and TV animation, education and training, and other fields.
AnimaCore
AnimaCore
Combined with deep learning technology, the self-attention mechanism based on the Transformer architecture enables seamless transition from speech and text to high-quality lip-sync animation. The core technology comprehensively analyzes the acoustic features of speech and the semantic information of text through multimodal learning to ensure that the generated lip-synchronization is synchronized with speech and can adapt to different language types and speaking styles.
- Support for private deployment
- Support for customized access to large language models
- Support for customized voice access
- Support for reasoning in national languages
- Support for customized training of animation models
MHC Talker
MHC Talker
Call AI animation engine API in Unreal Engine to drive Metahuman character high quality dialog in real time. One-click configuration of intelligent dialog characters, features include: text or voice dialog, real-time lip-sync driving, post-processing fine-tuning of expressions, voice re-reading natural head movement, body movement dialog logic driving, animation data recording, etc.
In addition, the DNA and BS modifications of character assets are particularly important to the character's animation performance, ensuring that the character's DNA and BS modifications conform to official MetaHuman standard specifications;
High-speed animated reasoning
Under the RTX 3090 GPU's standard operating environment, for animation generation or similar tasks, an average of at least 50 seconds of animation inference can be completed in less than 1 second, and at least 25 seconds of animation data can be output, demonstrating extremely high real-time processing capability.
ultra-low resource footprint
The model has been lightened and quantized to significantly reduce memory and video memory usage while maintaining high accuracy in the main task. It is suitable for most of the mainstream civilian GPUs, and can be adapted to different platforms (windows, linux) to achieve good resource utilization, and can run stably in both high-performance servers and general-purpose PC environments.
Intelligent segmentation mechanism
For long text or audio reasoning scenarios, the model can further improve the overall reasoning efficiency and significantly reduce the waiting time under the synergistic effect of segmented processing and caching mechanism.
Scalability and Deployment
Adopting modular and pluggable design concepts: the core reasoning and pre-processing and post-processing parts can be developed and maintained relatively independently. Distributed training and reasoning framework is able to handle massive amounts of data and expand at any time to meet growing business needs. Easy to integrate into existing production environments or quickly build new application scenarios.
Stability and Accuracy
Test and validate under multiple data distributions and complex scenarios to ensure stable and reliable results through error detection and robustness testing. Continuous monitoring and regular updating of model parameters, timely detection and repair of anomalies, so that the system still maintains good accuracy after a long period of operation.
Easy customizability
The model structure or reasoning process can be flexibly adjusted according to the actual business needs to balance the speed and accuracy requirements of different tasks.
Easy to maintain: perfect log monitoring, alarms and automated deployment system, can effectively reduce operation and maintenance costs and ensure rapid iteration.