Exit
  • Global community
    • Language:
      • Deutsch
      • English
      • Español
      • Français
      • Português
  • 日本語コミュニティ
  • 한국 커뮤니티
3

Exploring Animations Based on Modeler !

Explorer ,
May 06, 2024 May 06, 2024

If the following approach can be realized, voxel models could to a greater extent replace traditional models, often freeing creators from the need to perform topology work or unwrap UVs.

Currently, most 3D animations in the market heavily rely on traditional skeletal systems and quad face models. My conception, however, is an entirely new approach to joint solutions based on voxel models. At the heart of this system lies the introduction of "invisible universal joints," which act as pivotal components connecting limbs, enabling unprecedented degrees of freedom and flexibility. Specifically, each joint not only supports unidirectional and planar rotations but also allows for adjusting the distance at connection points within a certain range, thereby more accurately simulating the functionality of complex human structures like the shoulder joint.

### Design Considerations

- **Universal Joints**: Develop a series of customizable-position universal joints for precise control over every action of the model, enhancing the naturalness of animations.
- **Distance Adjustment**: Incorporate a mechanism for adjusting distances at joints to mimic limb extension and retraction, augmenting the realism of animations.
- **Interconnected Functions**: Drawing from the principles of parenting and IK (Inverse Kinematics) in skeletal systems, enable natural chain reactions between joint movements, boosting animation efficiency and fluidity.

### AI-Assisted Animation Optimization

To refine this system further, I propose integrating AI technology, specifically aimed at correcting any uneven seams or visual inaccuracies that may arise from joint manipulations. By training specialized AI models, we can instantly conduct seamless repairs after users adjust the model's actions, preserving the continuity and realism of the model's surface. Considering your company's leadership in technological innovation, I am confident that realizing this AI-assisted feature is well within reach.

### Accompanying Illustrations

To visually illustrate my concept, I have attached three key example images:
1. **Original Voxel Model**: Showcases a meticulously crafted voxel male model, serving as the base model.
2. **Sphere-Jointed Puppet in Pose**: This image depicts the model transformed into a sphere-jointed puppet, with joints set and posed in a specific action, offering a preliminary glimpse into its dynamic potential.
3. **Model After AI Refinement**: The final image exhibits the model after AI processing, where flaws at joints are impeccably corrected, resulting in a smooth surface and natural movement.

### Conclusion
Long live voxel models!p (1).jpgp (2).jpgp (3).jpg

Idea No status
TOPICS
Discussion , Question
235
Translate
Report
Community guidelines
Be kind and respectful, give credit to the original source of content, and search for duplicates before posting. Learn more
community guidelines
3 Comments
Adobe Employee ,
May 13, 2024 May 13, 2024

really interesting idea!  will digest with the team. Thank you.

Translate
Report
Explorer ,
May 14, 2024 May 14, 2024

Please consider this comprehensively; I'm merely offering an inspiration. The implementation could perhaps be simpler. Recent software updates have added many features but have become somewhat cumbersome in design, creating a slight discord with the original style. In fact, voxel sculpting can certainly handle details; it's just a different approach to sculpting. For instance, the original fill mode of the smoothing tool was quite good. I firmly believe that professionalism and user-friendliness can coexist. Additionally, I suggest incorporating an automatic display of specific dimensions, which would be very practical. The Boolean mode is inherently well-suited for creating parts.

Translate
Report
Explorer ,
May 19, 2024 May 19, 2024
LATEST

Based on my testing, the repair results are not good when there are perspective issues or incomplete characters in the images. Maybe we should directly use AI to repair voxel models? The preparation of the training dataset would be something like this:

First, gather a bunch of characters with rigged skeletons and pose them in different actions. Then, disassemble these characters at the joints and pose them in the same actions. Of course, all of them need to be converted into voxel models. Then, let the AI learn the differences between the two. The same method should also be applicable for adjusting facial expressions.

Translate
Report