← back to lab

Reusable prompt templates for logic-native video generation.

Canonical prompt scaffolds for omni-model, instruction-based editing workflows. Each template includes stage-by-stage instructions and compiled output.

2026-05-10omni-2026.05Beginner

Instruction-based Remix: Fiber City

A canonical prompt scaffold demonstrating native multimodal understanding through iterative instruction-based editing. This template showcases how to transform a static scene into a dynamic, physiologically synchronized environment while preserving physical consistency across all edits. The three-stage approach (base → modify → refine) ensures the model maintains causal fidelity without resorting to pixel-level substitution.

remixinstructionfiber-citydynamic
2026-05-11omni-2026.05Intermediate

Blackboard Derivation: Symbol Continuity

A comprehensive prompt scaffold for testing and validating symbol continuity, spatial anchoring, and logic-native layout planning in blackboard-style mathematical video generation. This four-stage template systematically evaluates the model's ability to maintain symbol identity across temporal occlusion events, plan spatial layout for multi-step derivations, and execute precise instruction following in symbolic manipulation tasks.

blackboardsymbolsmathcontinuity
2026-05-13omni-2026.05Intermediate

Character Consistency Across Takes

A robust template for maintaining character identity, clothing details, and anatomical proportions across multiple camera angles, temporal discontinuities, and action sequences. This four-stage scaffold systematically tests the model's capacity for identity preservation under challenging conditions: motion-induced deformation, camera angle transitions, and brief temporal gaps. Essential for validating character-grounded narrative video generation.

characterconsistencymulti-angleidentity
2026-05-22omni-2026.05Advanced

Conversational Editing: Multi-Turn State Preservation

A specialized template for testing the model's ability to maintain narrative state across multiple sequential editing turns in a conversational interaction. This four-turn scaffold validates cumulative state preservation, instruction-native context understanding, and the absence of regression to prior states during iterative refinement of the same video content.

dialoguemulti-turnstate-preservationediting
2026-05-28omni-2026.05Advanced

Material Physics Benchmark: Fragility and Rigidity

A comparative physics benchmark template designed to systematically evaluate material-specific behavior under identical impact conditions. This three-scenario scaffold tests whether the model maintains accurate physical properties for glass (fragile/shattering), metal (rigid/elastic collision), and fabric (compliant/deformation). Critical for validating physics-native understanding of material properties in video generation.

physicsmaterialsimpactfragility