Text-to-video lets you describe a scene from scratch, but for a lot of practical work there is a better starting point: an image you already like. Image-to-video modes use a still frame to define how the scene looks, and your prompt focuses only on how it moves. That split makes results more predictable. This guide covers the basics: when to use image-to-video, how to prepare the starting frame, and how to write the motion prompt.
Text-to-video or image-to-video?
Use text-to-video when you are still exploring what a scene should look like and are happy to let the model propose the visuals.
Use image-to-video when the look is already decided: a product photo, a character design, a campaign key visual or a frame from a mood board. The image anchors the subject, palette and composition, so your words can go entirely into action and camera direction.
Not every model supports image input, so check that the mode you need is available for the model you've selected before you prepare anything.
Step 1: Prepare a strong starting frame
The starting image does most of the work, so choose it carefully:
- Clear subject. One obvious focal point that the motion will center on.
- Room to move. If the camera will push in or the subject will turn, leave some space around it.
- Right shape. Match the frame's aspect ratio to the clip format you plan to publish.
- Clean details. Small artifacts in a still image tend to become more noticeable in motion.
If you don't have a suitable image, generate one first. Settling the look as a still, for example when you create the starting frame in an image generator, separates design decisions from motion decisions, so you aren't fixing both at once.
Step 2: Decide on one moment
Short clips have room for one clear event. Before writing the prompt, finish this sentence: "In this clip, the viewer sees ___." Good answers are single beats:
- The product rotates to reveal its label.
- Steam rises from the cup as morning light brightens.
- The character looks up from the map toward the mountains.
If your answer contains "and then," you probably have two clips.
Step 3: Write the motion prompt
With the visuals already defined by the image, the prompt should describe three things:
- The subject's action: what moves, and how.
- The camera: static, slowly moving closer, following, or panning.
- What must stay recognizable: details you don't want altered.
Example for a product photo:
The bottle rotates slowly to the right to reveal the front label. Static camera, centered framing. Keep the bottle shape and label colors unchanged.
Example for a character illustration:
She lowers the paper map and turns her head toward the distant peak. The camera moves slowly closer from behind her shoulder. Keep her green jacket and hairstyle consistent.
Nano Banana 3's image-to-video mode guidance follows the same logic. Give the viewer a focal point, choose an action with a purpose, and connect the camera to the action. It also suggests mentioning any visual details you want to keep recognizable.
Step 4: Check the generation settings
Before submitting, confirm:
- The input mode is image-to-video and your image is attached.
- The format matches where the clip will be published.
- The clip length and resolution suit the use.
- The displayed cost fits your budget for this attempt.
Video generations usually cost more than images, so a ten-second settings check is worth it.
Step 5: Review the whole clip
Watch the entire clip with your one-sentence intent in mind:
- Does the action read clearly from start to finish?
- Does the camera keep attention on the subject?
- Did any important detail change shape or color mid-clip?
- Is the pace right for where it will be used?
Fixing common problems
Too much happening. Reduce the prompt to one subject, one action and one camera direction. Remove secondary motion like passing people or changing weather.
The subject warps or drifts. State the details to preserve more explicitly, and check that the starting image shows them clearly.
The camera wanders. Specify a simple behavior, such as a fixed view or a slow push in, and say what it should stay on.
Nothing much moves. Name the action more concretely: "turns to the right," "lifts the cup," "door swings open."
Change one element per attempt so you can tell what fixed the problem.
Use rights-cleared images
The starting frame should be something you are allowed to use: your own product photography, images you generated, or licensed material. Be particularly careful with real people's likenesses.
Summary
Image-to-video works best when you treat it as two steps. First lock the look with a strong still. Then direct one clear moment of motion with a simple camera instruction. Keep prompts short and specific, check the settings before you submit, and review the whole clip against your intent.
Top comments (0)