Apple is fully participating in the AI ​​race. After first showing a model that can edit images, an AI model that converts still images into animations is now shown.

It’s been a while coming for Apple, but the iPhone maker has been fully embroiled in the AI ​​debates in recent weeks. A week ago, Apple-affiliated AI researchers demonstrated MGIE: a model that can manipulate images based on text prompts. Today Apple is releasing a new research paper on Keyframer: an LLM with animation capabilities.

From still images to moving images

The underlying principle of Keyframer is not much different from the model shown previously. You show the model an image (in .svg format) and describe in words what you want to see. The difference is that Keyframer is trained to convert a still image into a short animation.

These are more simple animations. One example that Apple describes in the article begins by depicting a planet against a background of stars. In this example, Keyframer can change the background color and make the stars twinkle. You can also have multiple designs created from the same animation. The paper provides further technical explanations of the LLM behind Keyframer.

AI in the iPhone?

It is still unclear what Apple plans to do with its LLMs. Tim Cook has already hinted that iOS 18 will be entirely dedicated to AI, without naming any specific features. The Google Pixel 8 and Samsung Galaxy S24 already have extensive AI capabilities, so the iPhone shouldn’t be left behind. Apple is expected to officially unveil iOS 18 during WWDC in June.