Create or transform an image
Start from text, or condition the result with up to three selected images for source, identity, style, environment, or prop.
Generation Production
Make the next frame.
Then make it move.
Start with Visual Lab references, then let Generation plan and produce the series. ReelGenerator shows the maximum cost and waits for approved planning before media generation.
Planning may call a provider; paid media waits for approved planning.
Source image Image result Frame-to-frame videoOpening + landing
“Carry the camera into the landing frame.”
One visible loop
Source, request, plan, result.
Every result keeps its inputs and place in the graph. Select it and continue in any direction.
- 01
Place your images
Import approved references or start with an empty canvas.
- 02
Ask in chat
Select up to three images and describe the next image or motion.
- 03
Review the plan
See exact inputs, operations, dependencies, and maximum media cost.
- 04
Approve the work
Durable workers publish immutable results beside their sources.
Current capabilities
Images in. Images and motion out.
Canvas is a Visual Lab for image and frame-to-frame video inputs. Generation owns the script, voice, series assembly, Final QA, and delivery.
Move between exact frames
Select one opening image and one landing image. Generate a silent 9:16 clip, four to fifteen seconds long, that connects them.
Keep the work in view