What if you could describe a short film and have an AI make it from start to finish, all on one infinite canvas? That's what xAI is now testing with Grok's new Imagine Agent Mode.
Multiple users on X have spotted the feature rolling out in beta on the Grok web interface. Dubbed "Imagine Agent Mode (beta)", it replaces the usual chat window with an infinite canvas — a blank workspace where you type a prompt and the agent progressively builds out images and videos.
https://twitter.com/testingcatalog/status/2049649752699249019
Screenshots show a new "Imagine" entry in Grok's sidebar. Tap it and you land on an open canvas with preset workflow templates: Create Worlds, Short Film, UGC Product Stories (user-generated content product packs), and Brand Identity. Describe what you want in one sentence, and the agent does the rest — generating and editing multiple images simultaneously, converting stills to video and auto-stitching them, trimming and crossfading clips, and exporting the final product.
According to TestingCatalog, the Imagine Agent can handle complex multi-step instructions like "generate a 1-minute short film" or "generate a full comic strip" or "generate a UGC product pack". Another user demonstrated the agent batch-generating three sets of product shots and three sets of model photos, then automatically combining them into a social media asset pack.
xAI hasn't made an official announcement yet. The company has been busy on the product front — it recently shipped Grok Voice Think Fast 1.0, a voice model now handling Starlink customer support. And the broader agent race is heating up: Ant Group dropped a trillion-parameter model designed for precisely this kind of fast, task-oriented execution.
