What problem does it solve? Emotion words like "happy" or "sad" give video models diffuse, unpredictable facial results. This Skill replaces emotion labels with FACS Action Unit codes (AU12 = lip-corner puller, AU6 = cheek raiser), giving muscle-level control over facial expressions in Seedance 2.0 generations — essential for forced smiles, mixed emotions, uncanny faces, and micro-performance in close-up dialogue. ## Core Features & Use Cases - AU Code Prompting: Write Action Unit codes directly into Seedance 2.0 prompts, either codes-only or codes plus short anatomical descriptions, with a full AU reference table grouped by facial region. - Emotion-to-AU Recipes: Standard EMFACS prototypes map emotions to AU combinations (Duchenne smile = AU6+AU12, anger = AU4+AU5+AU7+AU23), plus guidance for blending expressions. - Dialogue Facial Acting: Schedule one AU set per spoken beat, combined with the [AUDIO: Xs] lip-sync block, so every line gets pre-line, during-line, and post-line facial beats. - FACS Reference Sheets: Generate small labelled-grid images of only the planned expressions via GPT Image 2 or Nano Banana Pro for character identity consistency. - Use Case: Direct a close-up monologue where a character masks fear as reassurance — the mouth performs a forced smile (AU12) while the eyes leak terror (AU7), beat-synced to the spoken lines. ## Quick Start Ask the AI to write a Seedance 2.0 close-up prompt where a character's expression shifts across timed beats using specific FACS Action Unit codes, such as a Duchenne smile fading into sadness.
