What problem does it solve? Image-to-video generation with Agnes Video V2.0 (ti2vid mode) often causes the character to drift — hairstyle, facial features, or outfit change mid-animation. This Skill provides a structured prompt architecture and negative prompt vocabulary to keep the character consistent with the reference image. ## Core Features & Use Cases - Three-Layer Prompt Structure: Separates prompts into a Preserve layer (features that must stay identical), an Animate layer (minimal necessary motion), and a Scene layer (background, lighting, camera). - Character-Retention Negative Prompts: Ready-to-copy negative prompt word groups targeting face change, identity drift, and appearance mutation, plus general quality terms. - API Parameter Guidance: Covers Agnes Video V2.0 specifics such as the 8n+1 frame rule, public-URL-only image input, fixed seeds, and aspect-ratio locking to avoid automatic normalization. - Use Case: When animating an idol character image into a 5-second stage performance clip, apply the three-layer template with the character-retention negative prompt so the face and outfit remain identical to the reference image. ## Quick Start Use the img2vid-character skill to write a character-preserving ti2vid prompt for animating my reference image of a singer on stage.
