Image to video
Image-to-video starts with a visual anchor. Prepare a clear reference, decide what should move and describe the camera separately. Use HappyBloom to keep the image and its prompt together while you plan the shot.
Choose an image with a clear subject and enough room for the intended movement. Match its proportions to the output frame where possible. A compressed screenshot can introduce artifacts that a video model may carry into the motion.
Separate subject motion from camera motion. For a landscape, you might ask for drifting mist while the camera remains fixed. For a product, use a gentle orbit instead of moving both the object and the camera in several directions at once.
Upload images and video references to Your Space and save their prompts. Local items remain on the current browser until you explicitly sync them. Cloud file endpoints check account ownership; moving an item to Trash hides its record and is not a promise of physical object deletion.
Preserve the composition of the reference image. A gentle breeze moves the leaves, soft mist drifts in the background, and the camera makes a slow cinematic push-in.
Try this prompt →No. There is no live model connection. If you save a preview, your reference can be stored locally or uploaded to your signed-in cloud space, but it is not sent to a video generation provider.