VidVid's prompt words are composed of 25 independent fields, covering the six dimensions of director's intention, space, lens, character, sound and constraints. By filling it in according to the specifications, you can stably produce videos with a uniform style and high usability, without relying on luck.
AI video models have different sensitivity to different dimensions in a prompt. Splitting information into independent fields is easier for the model to follow accurately than one long paragraph.
Predictable image content, unified style expression, and measurable quality standards. Every field has clear priority, format, and checklist.
Cross-shot character faces, costumes, and tones do not drift. Triple mechanism ensures this: reference photo + character consistency + character constraint.
Fill in order P0→P3. When time is tight, ensuring P0 completeness is enough to output technically usable clips.
The 25 fields are organized into six layers from base to modifier, from space to time, from image to sound, following a progressive-overlay principle.
| Layer | Function | Field Count | Core Responsibility |
|---|---|---|---|
| Creative Intent | Style Governance | 1 | Establishes overall visual style as the decision benchmark for all fields |
| Picture Base | Base Parameters | 2 | Aspect ratio, resolution, quality tier, and other underlying technical parameters and global constraints |
| Space | Environment Building | 2 | Builds the physical space and lighting system, forming the 3D carrier of the image |
| Cinematography | Image Organization | 4 | Organizes visual storytelling through camera movement, composition, color, and depth |
| Character | Character Portrayal | 4 | Defines the character's external appearance and dynamic performance |
| Props | Prop Configuration | 1 | Arranges key items and handheld props in the scene |
| Quality / Schedule / Render / Transition / Audio / Narrative | Multi-dimensional Modifiers | 11 | Cross-layer guarantees for image quality, emotion, timing, connection, sound, and dialogue |
When the character limit is exceeded, cut in this order — P0 must never be cut, P1 is prioritized, P2 can be compressed, P3 is cut first.
| Priority | Level | Field Count | Consequence of Missing |
|---|---|---|---|
| P0 | Fatal | 12 | Video unusable or serious errors |
| P1 | Core | 7 | Video usable but quality significantly reduced |
| P2 | Enhancement | 4 | Quality meets standard but lacks professional polish |
| P3 | Optional | 2 | Does not affect main expression |