New inference
Upload images of a scene, describe a task, and get a predicted next-best viewpoint.
Take or select up to N photos. Each is center-cropped to 1024×1024.
Tap a thumbnail to set it as the start frame.
No images yet. Choose images or pick an example to begin.
Tap on the start frame to place a point.
No point set
Result
Open viewer in a new tabThe green camera frustum shows the predicted viewpoint for executing the instruction.