Gemini Video Prompt Generator

The Gemini video prompt generator sits right above this text. Fill in the fields, choose Plain text or JSON, copy the single prompt it writes, and paste it straight into Gemini.

Only four inputs are needed: the concept (your subject and setting), what happens in the clip, the mood, and the camera move. Sound or dialogue and the include or avoid field are optional.

For example, set the concept to “lighthouse at dusk” and what happens to “a beam sweeps over the waves.” The tool turns that into one clean prompt, camera and mood wording included, ready for the app.

Worth knowing before you start: the Gemini app now generates video with Gemini Omni 1.1 Flash, the model that replaced Veo 3.1 inside the app.

How to Use the Gemini Video Prompt Generator

Fill in the fields

Now to the fields themselves. Four are required: concept (your subject and setting), what happens in the clip, the mood, and the camera move. Two are optional: sound or dialogue, and include or avoid for anything extra you want in or out of the frame.

Three quick fills to show the range:

• Product shot: concept “skincare bottle on marble”, what happens “steam rises as the lid lifts”, mood “clean and calm”, camera “slow push in”.
• Travel clip: concept “narrow street in Lisbon at golden hour”, what happens “a tram passes and pigeons scatter”, mood “warm and lively”, camera “handheld follow”.
• Short story clip: concept “a lighthouse keeper at his desk”, what happens “the lamp cuts out and he looks up”, mood “tense and quiet”, camera “slow zoom on his face”.

Then choose the output format, Plain text or JSON. Either way you get one ready-to-paste prompt.

Copy the prompt into Gemini

Copy that prompt and paste it into the video option in Gemini. The app generates video from text, photos or both, and it needs a Google AI subscription, per Google’s video generation page for Gemini.

Not happy with the first take? Edit it by chat. Ask for a slower camera or a softer mood, and the app reworks the clip inside the same conversation.

Gemini Omni and Veo 3.1: What the Gemini App Uses Now

So what is running when you paste that prompt into Gemini? The model name changed, and that trips people up.

What Gemini Omni is

Google’s Gemini Omni announcement calls it a model that can create anything from any input, starting with video. Text, images, audio and video all go in. Video comes out, and you keep editing it through chat.

In the app, the model is named Gemini Omni 1.1 Flash. It is the one that replaced Veo 3.1 inside Gemini. That is the swap to keep in mind: your prompt goes to Omni now, not to Veo.

Where Veo 3.1 stands

Now the developer side, which is less tidy. The Gemini API video documentation lists gemini-omni-1.1-flash as the video model there too, with text, images, audio and video as input and multi-turn edits. The Veo 3.1 preview models on the Gemini API were shut down on 22 October 2026.

So the idea that Veo 3.1 simply continues for developers needs an asterisk. Google points people to two options: Gemini Omni Flash, or the GA Veo models on its enterprise platform. Only those GA Veo models carry on. The preview models in the Gemini API are gone. If you build on the API, check which of the two you are actually using before you rewrite anything.

What Each Field Adds to Your Prompt

Back to the tool itself. Each field covers a part of a good video prompt. Google’s Veo 3.1 prompting guide lists the parts like this: subject, action, style, camera position and motion, composition, focus and lens effects, and ambiance. The tool splits that list into four fields.

Concept and what happens

Concept is your subject and setting. “Lighthouse at dusk” says what sits in frame and where. What happens is action, and it is the field you should never skip. With no action the model has little to animate, so it invents something itself. “A beam sweeps over the waves” gives it movement to work with.

Mood and camera

Mood covers style and ambiance: the look, the light, the feel of the clip. Camera covers movement and framing. Vague words do little here. “Nice shot” gives the model nothing to follow. “Slow push in, low angle” gives it a move it can run.

Those two fields are what turn a flat scene into a shot. For text prompts instead of video ones, the Gemini prompt generator covers the rest of the app.

Sound, Include or Avoid, and JSON Output

Describing sound

Sound or dialogue is optional. Fill it in when the clip needs a specific audio layer, like soft background music or natural street sound.google/models/gemini-omni/” target=”_blank” rel=”noopener”>Google DeepMind’s Gemini Omni page, whose sample prompts spell out audio wishes like this. Leave the field blank and the tool stays quiet about sound, so Gemini picks what fits. One catch worth flagging: Google’s own pages disagree on audio output. The product page says videos get generated audio, while the launch post said audio output was not yet available at first. Features also vary by plan and region.

Include or avoid

Use this field for anything the other fields miss. Include covers extras you want in frame. Avoid covers what should stay out.

Plain text or JSON

Plain text writes the prompt as one paragraph. It pastes straight into Gemini with no cleanup. JSON keeps every field labelled, which helps when you store prompts or reuse them later. A JSON prompt generator can shape similar structured prompts for other tools. This is a format choice only. Pick whichever suits your workflow.

What Google Has Confirmed, and What Is Unclear

Some of this is settled. Some of it is not. Here is which is which.

Confirmed on Google’s side: app clips run up to 10 seconds, users must be 18 or over, and every video carries an invisible SynthID watermark. The Gemini Omni Flash model card also lists three honest weak spots: keeping full consistency through edits, complex motion, and perfectly accurate rendered text. In practice, text inside a clip may come out mangled, and busy movement can look wrong.

Sound stays murky. The two Google pages that clash over audio output are joined by that model card, published in May 2026 and updated that August, which describes the output as video with audio. The timing could explain the gap, since the launch post dates to the earlier moment. Google has not said so directly. So plan for sound in your prompt, but do not count on it.

One thing is missing: daily generation limits. Google’s pages do not state them. Anyone quoting exact numbers there is guessing.

The generator above stays narrow on purpose. For prompts beyond video, the AI prompt generator handles the rest.

Bottom Line

That is the whole workflow. Fill the four main fields: concept, what happens, mood, camera. Add sound or dialogue and include or avoid only when the clip needs them. Pick Plain text or JSON, then copy the one prompt it writes.

Paste it into Gemini and let the chat do the rest. Ask for a slower camera, a softer mood, or a tighter frame, and the app reworks the clip in the same thread. The tool gets you the first prompt. Everything after that is a conversation.