Write a clear Veo 3.1 product-scene prompt
By Drizzle team
2 min read
Separate the product reference, visible action, camera direction, and sound. A concise scene is easier to evaluate than a prompt full of competing effects.
Describe one scene and one movement
For a fictional tote: “The referenced canvas bag sits on a plain desk. The camera slowly moves closer to the inside pocket. Soft daylight, stable composition, quiet room ambience.” Add the product details that must remain unchanged.
Use the reference input supported by the selected mode. Do not expect a written brand name to reconstruct a small label accurately.
Keep settings out of impossible combinations
Google documents different controls for references, frame inputs, extension, resolution, and duration. Confirm the supported combination in the current Veo documentation. A prompt cannot override an endpoint’s input constraints.
When speech is needed, keep it short and approved. State which sounds belong in the scene instead of asking for a cinematic result with no audio direction.
Review before making the next prompt longer
Compare the ending with the source image. Look for altered labels, scale, handles, and hand contact. If the product changes, reduce motion or improve the reference before adding more adjectives.
Use the Veo prompt builder and model guide. These are external-model learning resources; the text builder does not generate a video or promise that Veo is selectable in Drizzle.