After a year of dedicated effort, my team and I are thrilled to announce the launch of MOJIKAI, a groundbreaking service that transforms textual descriptions into customizable 3D assets, avatars, and scenes.
What Sets MOJIKAI Apart?
Unlike other platforms that generate 3D content directly, MOJIKAI operates as an agentic system, leveraging a comprehensive database of parametric procedural templates. This approach allows users to create and fine-tune 3D models with remarkable precision.
How Does It Work?
When you input a description, MOJIKAI’s AI selects the most suitable template(s) and configures them according to your specifications, delivering the final 3D model in GLB format. For instance, you can request a fork with three or four prongs and adjust its length. If the template permits, you can even provide exact measurements, and the AI will adapt the model accordingly. Similarly, asking for an avatar of a young Japanese man will yield a model reflecting typical characteristics: Asian, male, mid-twenties, lean build.
Advantages of MOJIKAI’s Approach
- Intellectual Property Assurance: By utilizing our proprietary templates, we eliminate copyright concerns associated with 3D data sources.
- Efficiency in Expansion: Adding a new template to our database is immediate and cost-effective, avoiding the need for extensive data collection or retraining large language models.
- Model Optimization: Our models are streamlined, requiring less hardware and memory, facilitating smoother performance.
- Versatility and Quality: Each template can produce numerous variations, all designed with correct topology and UV maps. We can enhance or replace templates without retraining the entire system. Additionally, we support rigging with bones and skinning, features often absent in other AI-generated 3D models.
Current Limitations and Future Prospects
While MOJIKAI’s capabilities are extensive, it currently cannot generate 3D content beyond the scope of its existing templates. Expanding our library requires time, expertise, and effort.
However, we are actively researching and developing an LLM capable of generating parametric procedural templates. While this project is still in its early stages, we are confident that it can bridge the gap between our existing templates without compromising the overall quality of the output. At the same time, it will ensure that each template remains highly customizable. Success in this area will enable MOJIKAI to expand the range of 3D assets, avatars, and scenes it can produce.
Experience MOJIKAI
We invite you to explore MOJIKAI and experience its capabilities firsthand. Request an account at https://www.mojikai.com/.
As we continue to innovate, we believe that integrating specialized systems like MOJIKAI with large language models offers a more efficient path forward than relying solely on scaling existing models. But that’s a discussion for the next post.