Prompt Gemini Omni for image-to-video

Last updated: October 12, 2026

Gemini Omni animates an approved still into a short clip with sound in the same pass, and follows timed beats written into the prompt. Pick it when you want a quick, on-model motion test with audio, or an action that must land at a set moment.

Let the agent write it

You don't have to write any of this yourself. The Layer agent knows how to prompt each model it runs, from the same guidance this page is built on. Attach the still in chat, name the model and say what should happen, for example: Animate this with Gemini Omni Flash 1.1: the mage casts a fireball to the right, locked camera, one shot, 3 seconds, no music. The agent writes the prompt, shows you its plan, and runs it when you approve. Use the rest of this guide when you want to write or adjust the prompt yourself, and see Write a brief for the agent for how to brief it.

When to use it

Use Gemini Omni when you have the frame and need it to move with sound: ability casts, idles that break into an attack on cue, short trailer and social shots. Every clip comes back with audio, and second-by-second timing in the prompt is followed.

Clips are 16:9 or 9:16 and run 3 to 10 seconds. For other shapes or longer takes, try another model from the model picker on the same frame, or see Prompt Seedance for image-to-video. To generate a clip with no image at all, see Create a video.

Versions on Layer

Gemini Omni Flash 1.1

The version this guide's examples were made with. It takes a First frame and a Last frame, and renders at 720p, 1080p or 4K.

Gemini Omni Flash

Takes a first frame only, at 720p. In our test it read the same prompt the same way as 1.1, with the same motion and the character on-model, so a prompt moves between the two unchanged.

How to write the prompt

Camera and single-shot clause, then the action, then the style, then the sound:

Locked-off medium shot in a single unbroken scene with no scene cuts. The mage slowly raises his staff and the ember at its tip swells into a ball of fire, then he thrusts the staff forward and the fireball streaks out of frame to the right while his cloak billows back. Stylised painterly game art, colours unchanged. Sound: a low whoosh as the fire gathers and a sharp roar as it launches, no music, no dialogue.

Three or four sentences that each make one decision work better than a paragraph of cinematic adjectives. The first frame already sets the scene and the light, so the sentences go to what the camera does, what moves, the style to hold, and what you should hear. That prompt is the one in the clip above.

What works

Timed beats

Write second ranges to place each action: "[0-2s] The mage stands still, breathing, as the ember on his staff flickers. [2-3s] He snaps the staff forward and a fireball streaks out of frame to the right." In our test, on the same frame and seed, the mage held still for two seconds and fired in the last one, as written. Without the ranges, the cast started at once and spread over the whole clip. Plain words work too: "After 2 seconds, ...".

One unbroken shot

Gemini Omni cuts between several shots by default. For a game asset you usually want one take, so write "in a single unbroken scene with no scene cuts". In our 3-second test neither clip cut, with or without it, so short clips are less at risk; keep the clause anyway, since it costs nothing and longer clips have room to cut.

Camera first

Open with the camera: locked-off medium shot, slow push in, handheld, drone shot. Naming a move gets you that move.

Specific motion

Say what moves, in what order and how: "slowly raises his staff ... then thrusts it forward". A vague "make it move" gives the model nothing to follow.

Sound as direction

Every clip has sound. Name the effects and the ambience ("a low whoosh as the fire gathers and a sharp roar as it launches"), put any spoken line in quotes with the accent you want, and add "no music" or "no dialogue" when you want none.

Say what must stay

The model does not know which parts of the still were hard to get right. Name them: "His hood stays up, and his face, the staff and the castle stay exactly as in the still." On our frame the character stayed on-model without it, so treat it as insurance for the parts you cannot afford to lose.

Readable text

Gemini Omni can render lettering. Put the exact words in quotes and say where they appear: a wooden sign above the door that says: 'The Ember Inn'.

What to avoid

Expecting a negative prompt

There is no negative prompt. Put exclusions in the prompt as short sentences: "No dialogue.", "No extra sound effects.", "No text."

Starting from a small image

A low-resolution start image gives a softer clip. Upscale the still first, then animate it.

Settings that matter

First frame is the still the clip starts on; on Gemini Omni Flash 1.1, Last frame sets where it ends, and the model animates from one to the other, so describe the transition in the prompt. Duration runs from 3 to 10 seconds; find the motion at 3, then render the length you need. Aspect ratio sets 16:9 or 9:16 and the resolution. Sound is always generated, so there is no audio switch. Above the prompt, Camera, Staging, Look and Performance take reference sets for each.

Prompts to start from

Spell cast

Locked-off medium shot in a single unbroken scene with no scene cuts. The mage slowly raises his staff and the ember at its tip swells into a ball of fire, then he thrusts the staff forward and the fireball streaks out of frame to the right while his cloak billows back. Stylised painterly game art, colours unchanged. Sound: a low whoosh as the fire gathers and a sharp roar as it launches, no music, no dialogue.

Idle, then a timed attack

Locked-off medium shot in a single unbroken scene with no scene cuts. [0-2s] The mage stands still, breathing, as the ember on his staff flickers and his cloak stirs in the wind. [2-3s] He snaps the staff forward and a fireball streaks out of frame to the right. Stylised painterly game art, colours unchanged. Sound: soft wind, then a sharp roar as the fireball launches, no music, no dialogue.

Use it in Layer

  1. Add the approved still to a session's board.

  2. Click View on the image, then Turn into video. Create video opens in the side panel with the image as its First frame.

  3. Click the model name at the top of the panel and choose Gemini Omni Flash 1.1. The first frame stays in place.

  4. In the prompt box, write the camera and the single-shot clause, the motion, the style and the sound.

  5. Leave Duration at 3 seconds while you find the motion, and set Aspect ratio to the shape you need.

  6. Click Create video. The cost shows on the button before you click.

  7. When the clip lands on the board, click View to open it in the viewer and play it. Then change one thing in the prompt at a time and run it again.

Learn more

  • Generate and edit videos with Gemini Omni Flash (Google AI for Developers): the maker's prompt guide, including the default to several shots, the timing syntax, first and last frames, sound and plain-sentence exclusions.
  • Gemini Omni (Google DeepMind): what the model is built for, with examples of image, video and audio references turned into one clip.
  • Gemini Omni Flash 1.1 prompting (Runware): the five beats a prompt should cover in three or four sentences, the single-shot clause, and writing sound as direction.
  • How to use Gemini Omni Flash 1.1 (fal): worked image-to-video prompts that name what in the still must stay put.

Related