Describe a short scene or start with an image. Plan the action, camera movement, and sound, then choose your output settings and review the estimate before generating.
Gemini Omni Flash 1.1 creates video from a written scene or a starting image. Text gives you room to invent a setting and subject together; an image gives you an opening composition to build on. Begin with one useful moment: a product turning toward the light, a person completing an action, or a camera revealing a location. Describe what changes over the selected duration and where the shot should settle. A clear ending makes the result easier to evaluate and use in a larger edit.
A short video has limited time to introduce a subject, show movement, and reach a conclusion. Choose one action instead of several unrelated events. A hand can place a cup on a table; a cyclist can pass a storefront; a curtain can lift in a breeze. Explain the direction and pace when they matter. Review the contact between objects and the transition into the final pose. A convincing first frame does not guarantee that movement remains plausible throughout the clip.
Choose a stationary view, a slow push, or a tracking movement that supports the action. If the camera moves, describe what it follows and how far the scene should open up. A small movement can be enough for a product reveal or atmosphere shot. Combining an orbit, a zoom, and a cut in a few seconds makes the desired result harder to specify. Check the edges of the frame and background geometry as the viewpoint changes, especially around straight lines and reflective surfaces.
Make the intended sound part of the scene description. A café can have quiet room ambience and a cup touching a saucer; a street can have passing tires and distant voices. If dialogue matters, keep the line brief enough for the duration and identify the speaker. Listen to the whole result before choosing it. Exact pronunciation, timing, voice quality, and lip movement are not guaranteed by a prompt, and you may still need to adjust the soundtrack in an editor.
Think about where the clip will sit in a finished piece. A brief pause at the end can make a cut easier, while an uncluttered area can leave space for a title added later. Describe a clean composition instead of asking the model to produce long passages of exact text. Watch the opening and closing frames as well as the central action. Download the result and test it in your intended layout before deciding that it is ready to publish.
Image-to-video begins with a single source image. Use it when a product, character, or environment already has the composition you want. The reference gives the scene a starting appearance; your prompt should focus on what moves, how the camera behaves, and what should be heard. It does not guarantee that every feature stays unchanged across frames. Keep the original and compare it with the generated clip, particularly around faces, labels, outlines, and fine textures.
Use an image in which the important object or person is easy to distinguish from the background. Leave enough space around it for the proposed movement. A tightly cropped source may not give the shot room for a turn or camera shift. Check the framing before uploading, and use material you have permission to animate. An image that already contains unclear anatomy or distorted lettering will not necessarily be corrected when it becomes a video.
The starting image already supplies many visual details. Spend the prompt on the action, pace, camera, and sound that the still picture cannot show. For example, ask for a slow reflection moving across a metal surface while the camera stays fixed. If an important part should remain still, say so. Review the output for unintended movement in the background or changes in the subject’s shape. Start another request with a more focused direction when the first result drifts.
Choose a landscape or portrait frame around where the clip will be used. A landscape shot can show the surroundings; a portrait composition can keep a person or product prominent in a vertical placement. The frame changes how much of the scene is visible, so consider the crop before choosing it. Keep important details away from the outer edges when you expect to add captions or interface elements. Review the actual downloaded video in the target layout.
Version 1.1 offers 360p, 720p, 1080p, and 4K output options in this workspace. Lower output sizes can help you assess an idea with a smaller estimate. Larger files may suit a later presentation or editing step, but they do not correct an action or restore a changed face. The higher-resolution options include upscaling; do not treat the 4K label as a promise of native 4K detail. Check the result itself before committing it to a final project.
Choose settings around the action you want to show, then review the estimate before starting a new video.
Choose a duration from three to ten seconds and check the estimate for the selected output size. A longer clip or larger output costs more. Start with the time the action actually needs, instead of automatically choosing the longest option.
Preparing a prompt or draft does not generate a video. Every new generation has its own cost, including a revised attempt at the same idea. This page’s workflow starts from text or one image; it does not offer conversational video editing or automatic clip extension.
Use text for a new scene, or add one image when the opening appearance already exists. Check that the subject has space for the movement you want.
State the action, camera direction, and sound. Give the action a conclusion and keep spoken lines short enough for the selected duration.
Select the frame, duration, and output size. Check the current credits and generation availability before continuing.
Review the entire clip for motion, identity, audio, and unwanted changes. Save the result for your next editing step only when it fits the intended use.
Plans for image and video
Annual plans are billed upfront. The same plan credits refresh every 30 days and expire at the end of that window; unused subscription credits do not roll over.
Explore the workspace and prepare a first request before choosing a paid plan.
For occasional image and video experiments with a predictable monthly ceiling.
Subscriptions renew automatically at the selected monthly or yearly rate until canceled. Cancel future renewals in your account.
For creators who iterate every week and need room for several directions.
Subscriptions renew automatically at the selected monthly or yearly rate until canceled. Cancel future renewals in your account.
For sustained, high-volume creation and access to the highest available output quality.
Subscriptions renew automatically at the selected monthly or yearly rate until canceled. Cancel future renewals in your account.
Prices are in USD. Subscriptions renew automatically; credit packs are one-time purchases.
Compare plansNo image is needed for text-to-video. Image-to-video uses one starting image. Choose the workflow based on whether you want to invent a scene or animate existing visual material.
The workflows on this page create a new clip from text or a starting image. They do not provide video editing, continuation, or a conversation that remembers earlier generated clips. Use a separate editing tool to assemble finished results.
A larger output can suit a different delivery size, but it does not guarantee better movement, anatomy, lettering, or identity. Review those details separately. The higher-resolution options include upscaling.
Include a brief line or a few concrete sounds alongside the action. Identify who speaks and leave enough time for the line. Listen to the result; exact speech, synchronization, and a finished audio mix still need review.
Review the estimate shown for the duration and output size you selected. Check it again after changing either setting. Preparing a draft does not itself spend generation credits; a new generation or revision is a separate request.
Choose one action worth showing. Add a starting image if it helps, describe the camera and sound, and review the settings before continuing.
