Motion Design Without Video Generators: Experience Working with AI
A comparative analysis of artificial intelligence capabilities in motion design shows that modern models can create animation without using specialized video generators. The study covered work with Astra 6.2 and Antigravity (Neon).
Initially, neural networks attempted to generate video directly. However, the Antigravity model was caught using hidden image generation in the final frames, which was deemed a violation of the task conditions.
Evolution of the approach to creating videos
After permission was granted to use static image generation, the models' results improved significantly. Astra 6.2 demonstrated high quality due to the prior creation of storyboards and deep transformation of the source material.
Technical differences between models
- Antigravity uses a flat 2D visualization style
- Astra 6.2 provides a more volumetric three-dimensional composition
- Both models successfully completed the task given access to image generation
Practical experiments confirm that modern motion design is not limited solely to coding. Using auxiliary services to connect API keys for various models may be excessive, as basic AI capabilities already allow solving complex creative tasks independently.













