Define the output objective
Start by deciding what the Google Omni Flash generation must achieve. Describe the output type, audience context and the single most important visual result. Keep this objective short because it acts as the priority reference for the rest of the prompt.
Lock the primary subject or scene
Describe the identity-critical elements that should remain stable. For Google Omni Flash, this may include a product silhouette, character wardrobe, environment anchor, material language or a consistent scene state.
Add camera, composition or movement logic
Choose only the camera or composition instructions required for the shot. Make movement physically readable and avoid stacking conflicting directions. If the workflow is image-first, use viewpoint and framing instead of unnecessary motion vocabulary.
Build lighting and atmosphere
Lighting should support the subject and creative intent. Describe direction, softness, contrast and practical sources. Atmosphere such as haze, rain, particles or reflections should reinforce the scene rather than hide it.
Apply style and constraints last
Add stylistic finish after the visual structure is clear. Use negative constraints only for likely failure modes. A shorter, prioritized restriction list is usually easier to audit than a long generic negative prompt.
Review, generate and revise one variable at a time
After the first output, identify the one variable responsible for the largest gap: subject, composition, camera, motion, lighting, style or restriction. Revise that layer first so you know which change produced the improvement.
Reusable Google Omni Flash prompt architecture
Primary anchor: define the main subject or scene
Camera / composition: one clear viewpoint and movement plan
Environment: context, depth and spatial relationships
Lighting: motivated direction, contrast and atmosphere
Style / finish: visual treatment and output target
Constraints: only model-relevant failure modes