Gemini Omni video generator: text and images to video

Text to video is the simplest way in. You describe a single scene in plain language and the model returns a short clip. There is no timeline, no node graph and no template to pick, so the only skill involved is describing what you want to see.

Strong prompts read like a shot list entry rather than a story. Name the subject, say what it does, describe how the camera moves and set the light. A weak prompt says "a beautiful cinematic video of a city". A strong one says "a rain-soaked neon street at night, a cyclist rides toward the camera, low tracking shot, reflections on wet asphalt". The second gives the model something concrete to follow and gives you something concrete to change on the next attempt.

When a prompt alone is too loose, attach images. The generator accepts up to 7 PNG, JPG or WebP files, up to 10 MB each, and uses them as visual anchors for the clip. Typical references are a product photo that has to stay on brand, a character sheet, a location still, or a mood board frame that sets the colour palette.

Use references for the things you cannot describe precisely and the prompt for the things you can. A face, a logo shape or a specific fabric is easier to show than to write. Camera movement, timing and mood are easier to write than to show. Splitting the job this way gives better results than either input alone and wastes fewer credits on rerolls.

Pick the aspect ratio first, because it changes how you frame the shot. 9:16 suits phone-first channels such as Shorts, Reels and TikTok. 16:9 suits landing page heroes, YouTube and presentations. Changing it later means regenerating the clip, not cropping it.

Resolution and length decide the cost. 720p and 1080p cost the same, so there is no reason to render 720p for a final cut. 4K costs more per clip and is worth it only when the clip will be shown large or cropped in editing. Length runs from 4 to 10 seconds; a 4-second clip is the cheapest way to test whether a prompt works, and a 10-second clip is better once you know it does.

Generation is billed in credits, not in minutes or subscriptions you have to guess at. The table shows exactly what one clip costs; it is calculated from the same configuration the generator charges from. A new account starts with 300 credits, which is 2 clips at the default 6-second length or 3 clips at 4 seconds.

Most wasted credits come from changing everything between attempts. A cheaper routine is: write a short prompt, generate a 4-second draft, look only at composition and motion, and ignore small details. If the framing is wrong, change the camera sentence. If the subject drifts, add a reference image. Only when the draft looks right do you render the final length and resolution.

Failed generations are refunded to your account automatically, so a technical error never costs you credits. A result that simply is not what you wanted still uses the credits, which is why drafting short and cheap first matters.

A generated clip is not the end of the process. If you like the shot but want a different mood, lighting or setting, the video editing mode takes your clip back in and restyles it while trying to keep the movement. For longer pieces, generate each shot separately, keep the same reference images so the look stays consistent, and cut them together in your usual editor.