Character Expression Director · публичный навык Open Skill
Character Expression Director
Character Expression Director — публичный Open Skill на Digen. Откройте, чтобы увидеть, что создаёт этот процесс.
ShortDrama
12 3 8
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1856918/2026-09-03/4zq7d4pn.png
# Character Expression Director
## When to use
- Activate **whenever we generate any character-driven image or video** where a character's face is visible: film/drama shots, short videos, MVs, dialogue scenes, emotional close-ups, portraits, reaction shots.
- Use standalone (the user describes an emotion/situation and wants a character shot) **or** as an overlay on other generation workflows (movie-maker, short-maker, music-video-lovagen, etc.): the same expression rules apply to every prompt that contains a visible character face.
- Do not use for content without characters, or when the user explicitly asks for a neutral/emotionless face.
## Core principles
1. The emotion must be **clearly visible on the face**, yet **natural and believable** — never exaggerated, theatrical, frozen, or mask-like.
2. The face must **never stay stiff or emotionless**. Always include subtle facial movement: eyebrow motion, blinking, eye direction, small mouth movements, cheek tension, forehead tension, and slight head movement.
3. Expressions must feel **cinematic, emotionally believable, and synchronized with the mood and action of the scene** — the same emotion reads differently in a quiet moment vs. an intense action beat. Adjust intensity to the situation.
## Emotion → expression map
Use these descriptors when composing prompts:
- **Happy**: relaxed eyes, a slight smile, lifted cheeks, an open, positive expression.
- **Angry**: sharper eyes, lowered eyebrows, facial tension, a more intense look.
- **Afraid**: wider eyes, a slightly open mouth, nervous tension, an uncertain expression.
- **Sad**: softer eyes, lowered facial energy, subtle tension around the mouth, a vulnerable look.
- **Surprised**: the eyes and mouth react visibly (widened eyes, raised brows, parted lips).
- **Determined**: a focused gaze, calm facial control, a confident expression.
- **Blends**: combine descriptors proportionally (e.g. sad + determined = soft vulnerable eyes with a firm, focused jaw).
## Prompt building blocks
- **Positive frame**: "The character's facial expression clearly shows {emotion}: {descriptors from the map}. Subtle facial movement: {micro-movements}. The expression is natural, believable, cinematic — never exaggerated, never stiff or emotionless — and synchronized with the {mood/action} of the scene."
- **Micro-movement set** (choose 3–5 per shot): gentle eyebrow motion, natural blinking, purposeful eye direction, small mouth movements, subtle cheek tension, faint forehead tension, slight head movement.
- **Negative guardrails** to avoid in the prompt: blank stare, frozen/expressionless face, plastic or doll-like look, overacted grimace, mask-like rigidity.
- **More examples**: ready-to-paste prompts for every emotion (image and video) live in `references/example-prompts.md`, including blends and intensity scaling.
## Steps
1. **Read the scene's emotion.** Determine the character's current emotion (happy / angry / afraid / sad / surprised / determined, or a blend) from the user's request or script, plus the situational context (quiet moment, confrontation, shock, loss, resolve…). Note the intensity level (subtle → strong) so the expression stays believable for that exact moment.
2. **Compose the prompt with expression direction.** Add the matching descriptors from the emotion map, the micro-movement set, and the scene-sync note. Example for a sad shot: "The character's face shows sadness: softer eyes, lowered facial energy, subtle tension around the mouth, a vulnerable look. Subtle movement: slow blinking, gentle downward eye direction, small mouth movements, faint brow shifts, slight head movement. Natural, believable, cinematic — not exaggerated — synchronized with the quiet, sorrowful mood of the scene."
3. **Generate.** If acting as an overlay on another workflow, inject the composed prompt into that pipeline's generation step (image or video) and keep its channel choices. If generating standalone:
- Images: `generate_image` with `mo