Apple has unveiled a new prototype generative AI animation tool called Keyframer which enables users to add motion to 2D images simply by describing how they should be animated.
A 31-page Apple research paper published on Feb. 8 explains that large language models (LLMs) are “underexplored” in animation despite the potential they’ve shown across other mediums like writing and image generation. The LLM-powered Keyframer tool is offered as one example of how the technology could be applied in the future.
Keyframer users can upload the image, type a simple instruction like “make the stars twinkle,” in the prompt box, and hit generate. Examples show how a Saturn illustration can move between background colors, or show stars fading in and out of the foreground.
Using OpenAI’s GPT4 as its base model, Keyframer can use OpenAI’s GPT4 as its base model to take Scalable Vector Graphic (SVG) files (an illustration format that can be resized without interfering with quality) and deliver CSS code to animate the image based on a text prompt.
The system can produce multiple animation designs in a single batch, and adjust properties like color codes and animation durations in a separate window. Keyframer uses a description-based method, thus automatically converts these changes into CSS, though the code itself is also fully editable. This approach is described as much simpler than other forms of AI-generated animation, which often need various applications and some coding experience.
Keyframer isn’t publicly available yet, and Apple’s user study only used 13 people, who could only play around with two simple, pre-selected SVG images when applying the tool. Apple also mentioned the system’s limitations in the study, specifying that Keyframer focuses on web-based animations like loading sequences, data visualization and animated transitions. Currently, the animation we all see in movies, TV shows and video games is too complex to produce using descriptions alone.
Keyframer is one of several generative AI programs that Apple has announced in recent months. Last year, the company introduced Human Gaussian Splats (HUGS),which creates animation-ready human avatars from video clips. Last week, Apple also released MGIE, a new AI model that can edit images using text-based descriptions.
For more info, visit appleinsider.com
Sources: Appleinsider.com, 9to5mac.com, venturebeat.com, theverge.com
Win 'The Art of DreamWorks Dog Man'! 
