Underwater city world model prompt
Copy this underwater city world model template. Find source guidance, a simpler variation and motion checks for your video model.
Combine an unusual environment with a simple navigation structure.
Copyable prompt
A submerged stone city with a wide avenue between intact buildings and a large statue ahead. Eye-level exploration with clear blue-green water and gentle overhead light. Sparse drifting particles, stable avenue geometry and open routes for slow navigation. Keep major landmarks recognizable as the viewer turns.
This is an editable starting template. The model, reference input and serving profile determine the output; evaluate the generated result before using it in a product.
Choose the input
A scene showing the avenue, ground plane and distant statue clearly despite the underwater lighting.
Describe a navigable scene, then use supported movement and look controls to explore it. Pair the prompt with a seed image that shows a clear route and stable perspective.
A simpler variation
Reduce particles and increase visibility before evaluating turns and reverse movement.
What to inspect
Revisit the statue from another angle and inspect building identities for discontinuity.
Record the input, supported control, model profile and the visible result. Change one variable at a time to learn which instruction affects your scene.
Models and next steps
Browse interactive world model prompts, or assemble your own with the prompt generator.
FAQ
Can I copy and edit this video prompt?
Yes. Copy the template, replace the subject and scene details, and adapt it to your model's supported input. Start with the simpler variation if the output becomes unstable.
Does this template guarantee the described result?
No. It expresses the intended scene or edit. Inspect the model's output for identity, motion and continuity before publishing or integrating it.
How should I evaluate a change to the prompt?
Keep the source input and serving profile constant, change one instruction, and compare the visible result over several runs. Measure responsiveness separately from output quality.
