Prompt patterns across Veo models
Describe the subject, action and camera move first. With an opening image, specify what should move and what should remain. Keep the same brief when comparing models, and change one instruction at a time.
Veo 3.1 and Gemini Omni Flash 1.1 examples, with prompts, inputs and settings to inspect.
No videos are available in this gallery yet.
Compare Google video examples from Veo 3.1, Fast and Lite alongside Gemini Omni Flash 1.1. Open a result to inspect its prompt and settings, then adapt the example in your workspace.
Veo and Gemini Omni are distinct Google models grouped in this gallery. Keep the model shown on each example as your reference. Use the linked model pages and comparisons to choose a workflow, then check the current quote before generating.
Describe the subject, action and camera move first. With an opening image, specify what should move and what should remain. Keep the same brief when comparing models, and change one instruction at a time.
Compare motion, framing and subject continuity across the whole clip. Veo and Omni do not share every input or editing option: check the selected model and mode before supplying images, references or a source video. One successful example does not guarantee the same result from a new prompt.
The example detail page shows the current price estimate for the settings shown. Your next render uses the current quote for its model, inputs, duration and resolution. Reuse the available settings, add your own sources where needed, and review the quote before submitting.
This gallery groups Google video examples, including Veo and Gemini Omni. They remain distinct models: inspect the name on the example and use the Omni versus Veo comparison to choose the right workflow.
Open the example detail page to review its prompt, settings and current price estimate, then use the recreation action. Check the selected model, supply any required source media and review the new quote in the workspace before generating.
Use the same subject and creative goal, then compare the whole result and the current quote. Keep the input mode explicit; reference and editing controls vary by model, so a preset should not be assumed to transfer unchanged.