character-intake

# Character Intake Produce a factual handoff; do not design or render. This node handles dual inputs: 1. **Image Input (Mode 1)**: Real person photo/selfie or existing cartoon/stylized character image. 2. **Non-Image Input (Mode 2)**: Local project/account/folder path, web link, or text brief. ## Input Contract Accept conversation history plus any local photo/image path, local project/folder path, URL, attachment, screenshot, or 1–3 sentence character/persona description. Read accessible materials before asking anything. ## Procedure 1. **Dual-Input Identification & Extraction**: - **If Image provided**: Extract facial topology, hair silhouette, signature facial marks (glasses, mole, freckles), and core temperament/demeanor (e.g. sharp geek, elegant scholar, vibrant youth, sassy rebel). Select only 3–5 recognition-bearing anchors for `identity_lock`. Do NOT lock the person's photo clothing. - **If Non-Image provided**: Extract character persona, profession, age group, core personality, brand/account context, and signature visual anchors. Build the `identity_lock`. 2. **Normalize Constraints**: - Record `must_keep`, `must_avoid`, `requested_exaggeration`. - Separate facts from assumptions. - Write `photo_attributes_released`: source photo clothing (ordinary t-shirts, casual wear, pajamas), background, lighting and casual pose are 100% released and discarded for bespoke redesign. - Record `inferred_preferences`: an evidence-based analysis of the person's aura, temperament, and likely fashion/art taste (e.g., techwear, cyber, luxury minimalism, haute couture, guochao, pop art) to inspire bold A/B/C wardrobe and style matching. 3. **No Unnecessary Friction**: - If materials are readable, do not ask the user to choose a head-to-body ratio (PG) or rendering style (MS); those are director decisions unless explicitly constrained by user overrides. - Ask only one concise question when unreadable or missing information would block execution. 4. **Handoff**: - Return `input_mode`, `identity_lock`, `project_context`, `confirmed_brief`, and `next_stage: direction_planning`. ## High-art photo policy For a photo input, the default reference mode is **identity anchor**, never portrait replica. Extract only factual recognition locks such as face topology, hair silhouette, demeanor, and distinctive facial markers; completely discard the casual photographed clothing and room background. Record an explicit high-likeness request only when the user asks for high likeness or a near-realistic portrait. Otherwise hand off a requirement for materially stylized output with bold, bespoke wardrobe tailoring, and flag portrait replication or plain photo clothing as a review failure. ## Output Contract Return normalized envelope: - `input_mode`: `"image"` | `"context_text"` | `"hybrid"` - `identity_lock`: 3–5 recognition-bearing facial, hair, signature and demeanor anchors only (no casual photo clothing) - `photo_attributes_released`: list of source-photo attributes (outfit, setting, pose) released for complete redesign - `inferred_preferences`: temperament analysis and inferred aesthetic fashion/art taste for A/B/C styling - `confirmed_brief`: normalized brief - `next_stage`: `"direction_planning"` (or `"character_intake"` if blocked)