The Grok Imagine Video 1.5 agent is now available, and it runs on Grok Image 2.0. The updated Imagine Video agent plans and connects shots in new and more diverse ways, delivering a more cohesive story, but it isn’t a complete refleshing of the older 1.5 Video launch from back in June 2026.
Readers could already generate clips with Grok Imagine Video 1.5. This new capability doesn’t change that, but instead changes the layer between the user and the actual imagine video generation model. It uses Image 2.0 to better portray a story, and keep multiple shots connected with greater continuity.
Grok’s X account posted an 11 second video that shows that cohesive nature in action, with the terrain and dragon not drastically changing frame to frame.
What Continuity Really Means
The update provides a greater sense of continuity across video. If you’ve previously watched AI-generated video, you can easily find where continuity drops and the AI model forgets the previous scene or information. That gap can be jarring, and is evident of AI-generated video.
What Imagine 1.5 with Agent 2.0 provides is a way to stitch those shots together. Now, the Agent 2.0 provides shot planning ahead of time, and then uses Imagine 1.5 to connect the shots together. This storyboarding method makes a more cohesive video, and is similar to the same techniques applied in real-world film development.
Image 2.0 Does the Planning
Imagine 2.0 has already had great changes to better keep it on track when generating multiple images of the same character or location, bringing greater consistency. Now, that same model is shifting from just still images to providing the planning logic behind Video 1.5.
When you describe the video, Imagine 2.0 takes the content, storyboards keyframes, and then Imagine Video 1.5 generates the content in between each keyframe. That provides consistency of content between each section, and doesn’t risk overflowing the smaller context windows that video models operate inside in.
No API Changes
Developers should not treat this as a new API model. The docs still list grok-imagine-video-1.5 as the video model, priced at $0.080 per second of video generation. There is also no new agent model ID, so this is really just an update to the existing product.
That means the change to the agent layer is invisible for both users and developers, except for the updated video that it now generates. Developers don’t need to make changes to their existing services that use Imagine Video 1.5, and users won’t see a difference in the primary user-facing interfaces on Grok and X.
