Keyframer: Apple’s Latest Generative AI Project
By Jess Weatherbed, a news writer specializing in creative industries, computing, and internet culture, with a background at TechRadar covering news and hardware reviews.
Apple researchers have introduced Keyframer, an innovative generative AI animation tool designed to facilitate the animation of 2D images based on textual descriptions.
In a recent publication on February 8th, Apple highlighted the untapped potential of large language models (LLMs) in the realm of animation, an area where their effectiveness has been demonstrated in fields such as writing and image creation. Keyframer, powered by OpenAI’s GPT4, serves as a prime example of how this technology can be leveraged.
By utilizing Scalable Vector Graphic (SVG) files, which allow for seamless resizing without compromising quality, Keyframer can interpret text prompts to generate CSS code for animating images. Users simply upload an image and input commands like “make the stars twinkle” to trigger animations such as color transitions in a Saturn illustration or the fading in and out of stars.
Illustrative frames created using Apple’s Keyframer prototype, featuring a simplistic depiction of Saturn and a rocket.
Keyframer enables users to generate multiple animation variations simultaneously and adjust parameters like color schemes and animation durations through a user-friendly interface. While no coding knowledge is required due to Keyframer’s automatic conversion of user inputs into CSS, advanced users have the option to directly edit the underlying code. This intuitive, description-driven approach contrasts with traditional AI animation methods, which often involve multiple software applications and coding proficiency.
Visual excerpts showcasing Apple’s Keyframer animation tool, including code and property manipulation windows.
A professional motion designer involved in Apple’s research expressed both excitement and apprehension about the tool’s potential impact on the industry. While acknowledging the transformative capabilities of such technologies, the designer emphasized that tools like Keyframer should be viewed as enhancements rather than replacements for human creativity. Embracing these innovations can ultimately elevate the skill set of animators, offering new avenues for creative expression.
Despite its promising features, Keyframer remains in the developmental stage and is not yet available to the public. The limited user study included in Apple’s research paper involved only 13 participants who explored the tool’s functionalities using predefined SVG images.
Apple’s research paper also outlines Keyframer’s current scope, emphasizing its focus on web-based animations such as loading sequences, data visualization, and animated transitions. Complex animation tasks found in movies and video games exceed the tool’s capabilities, highlighting the ongoing challenges in generating intricate animations solely through descriptive inputs.
Keyframer represents Apple’s latest endeavor in the realm of generative AI, following recent announcements such as Human Gaussian Splats (HUGS) for creating animated human avatars from video clips and MGIE, an AI model for image editing based on textual descriptions.
