Dolly In
Camera moves forward toward the subject. Creates intimacy and focus.
e.g. "slowly dolly in toward the face"
Image to Video
Cosmos 3 Super Image to Video
Upload any image and describe how it should move. The Generator Tower produces temporally consistent video clips with strong controllability over camera motion, physical dynamics, and scene behavior.
Your generated video will appear here
Upload an image and click Generate
Camera Guide
Camera Motion Language for Image to Video
Use cinematography terms in your motion prompt to give the Generator Tower precise direction.
Dolly Out
Camera pulls back from the subject. Creates reveal or isolation effect.
e.g. "dolly out to reveal the full scene"
Orbit / Arc
Camera circles around the subject. Great for 3D product showcases.
e.g. "smooth orbit left around the object"
Tracking Shot
Camera follows the subject laterally. Good for motion and action.
e.g. "tracking shot following the car"
Tilt Up/Down
Camera tilts vertically. Good for establishing height and scale.
e.g. "tilt up to reveal the skyscraper"
Crane Shot
Camera rises or descends. Creates dramatic reveal from above.
e.g. "crane shot lifting above the forest"
FAQ
What image formats can I upload?▾
Cosmos 3 Super accepts PNG, JPG, and WEBP images up to 10MB. For best results, use a clear, high-resolution image with a well-defined subject and good lighting.
How does the Generator Tower animate my image?▾
The Generator Tower uses a diffusion-based process conditioned on the Reasoner Tower's understanding of your image's physical properties. It generates future frames that are physically consistent with the original scene.
How long can the generated videos be?▾
You can generate 3, 5, or 7-second videos. 5 seconds is the recommended default for most use cases. Longer videos use more credits.
What resolution are the output videos?▾
Videos are generated at 720P quality. Landscape is 1280×720, portrait is 720×1280, and square is 720×720.