Grok Imagine opens Agent Mode beta on web with infinite canvas

• bennash's repost of XFreeze said Grok Imagine Agent Mode beta had gone live on the web, framing it as a creative agent that works on an "infinite open canvas."
• In venturetwins' froyo brand demo, the agent handled prompt writing and canvas buildout inside one session, which is a noticeable shift from one-shot image generation.
• icreatelife's screenshot showed the current beta pitch in plain language: brainstorm with the agent, generate and edit images, turn them into videos, then stitch those videos together in one place.
• An Official Grok Imagine post already advertised longer videos at 720p, while TestingCatalog's rollout report said the new canvas agent was appearing for Grok web users before any formal xAI announcement.

You can browse the official Imagine upgrades post, open a live Grok Imagine template, and compare that with TestingCatalog's earlier template report, which listed photo-to-video, photo style edit, and photo edit video as the first workflow types.

The clearest change is interface shape. bennash's repost described Agent Mode as a full creative agent on an infinite canvas, and icreatelife's screenshot showed the beta swapping a normal chat box for a workspace with image, video, resolution, duration, and aspect-ratio controls in the same panel.

TestingCatalog's rollout report said xAI had started rolling the feature out on Grok web without a matching official announcement. That makes this feel more like a quiet product surfacing than a polished launch post.

Building a froyo brand inside Grok Imagine Agent

In venturetwins' demo, the input was just "I want to build a froyo brand," and the agent expanded that into brand elements on the canvas. The useful part is not the final mockups, it is that Grok is now absorbing the annoying middle layer of creative tooling, prompt drafting, layout assembly, and iteration inside one run.

The beta card in icreatelife's screenshot lists the sequence explicitly:

• brainstorm with the agent
• generate images
• edit images
• turn images into videos
• stitch videos into longer videos

That lines up with the official Imagine upgrades post, which already advertised longer videos at 720p. The new piece is the agentic wrapper around that media stack.

Before Agent Mode appeared, carolletta's template post was already linking straight into a public Grok Imagine template, the Elon Wins template, which suggests xAI had been building reusable creative flows around Imagine rather than treating it as a blank prompt box.

TestingCatalog's template report said the first custom template categories were:

1. Photo to Video
2. Photo Style Edit
3. Photo Edit Video

That same report said an Image Reference template type was in the works. Agent Mode looks like the next layer up: not just reusable prompts, but a persistent canvas that can chain those media steps into one creative session.