Multi-View Consistent Character Generator (HQ Realistic) · Open Skill 공개 스킬
Multi-View Consistent Character Generator (HQ Realistic)
Multi-View Consistent Character Generator (HQ Realistic)은 Digen의 공개 Open Skill입니다. 열어 보면 이 워크플로가 무엇을 만드는지 알 수 있습니다.
Custom
5 0 3
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4173183/2026-10-15/xteg65vc.jpg
# Multi-View Consistent Character Generator (HQ Realistic) One deliverable: a **single square, HQ photorealistic** image with a **3 rows by 3 columns** contact sheet (**9 equal panels**) showing **one character** from **9 distinct camera positions and lens feels**. Identity, hairstyle, clothing, props, and environment stay **consistent**; only **camera position, angle, and framing** change per panel (small natural head or hand adjustments allowed when the angle requires it — never a wardrobe or identity swap unless the user explicitly asks to refine specific elements on a follow-up). **Positioning:** efficient character-sheet editing for **advertising layouts**, **photography mood boards**, and **creative enhancement** — not a generic one-off portrait. Prompt assembly and the default 9-view roster live in `references/multi-view-grid-template.md`. Read that file before every generate or redo. No structured project state. Progress is conversation plus the canvas asset. ## Global language lock Handbook (this file and all `references/`) is **English only — zero Han/CJK**. User-facing copy uses locked **`user_lang`** per `references/00-language.md`. Image `prompt` is **always English**. ## Pause + guidance (hard) **Every time you stop and wait for the user**, call **`set_guidance` in the same turn** as the last user-visible message — **no exceptions**. Pauses include: optional intake, policy refusal, explain-the-layout without generate, recoverable failure, and **after every successful image**. Before any pause turn, `read_skill_file` → `references/guidance-buttons.md` and pick **2–4** actionable buttons for that situation. - **Forbidden:** ending a wait turn with only prose and no `set_guidance`. - On the `set_guidance` turn: **no other tools** in that turn. - **Forbidden:** calling `set_guidance` in the same turn as `generate_image` or `recognize_image`. ## User-visible speech (hard) The user must not see the pipeline. Tools still run; chat text does not narrate them. **Never tell the user:** that you are reading a skill or reference; todos; tool names; channel names; lock block names; the full prompt; `entity_type`; silent QC or redo. **Allowed user-visible text:** - Optional intake: one short note when the user gave **no character description, no reference upload, and has not tapped direct generate or asked to generate** — offer defaults via buttons. **Stop** with `set_guidance`. - Starting generate: at most **one** short line in **`user_lang`**. Then tools only until the image is ready. - Success: one or two sentences in **`user_lang`**. Same turn: **`set_guidance` only**. - `success=false` when user action can help: short failure note + **`set_guidance`** in the same turn. - Refuse: short polite refusal + **`set_guidance`**. **Direct generate is first-class.** If the user taps **direct generate**, **generate without upload**, or equivalent, **do not** require an upload — proceed with **T2I** using defaults in `references/guidance-buttons.md` unless the same message already specifies character, pose, or scene. If the user already asked to generate (with or without a photo), **do not** run a long questionnaire — read the template, infer missing slots from the reference or sensible defaults, and generate. ## When to use - User wants a **multi-view character sheet**, **3x3 angle grid**, **consistent character across cameras**, or **lens / angle exploration** for **one subject**. - User uploaded a **character photo** and wants **the same person** from many angles while **keeping outfit and scene**. - User wants **direct generate** / **no upload** — describe or default an **adult fictional character** and produce the full 9-panel sheet. - Follow-up: **regenerate**, **wider vs tighter lens mix**, **more extreme high/low angles**, **numbered panels on/off**, **match reference scene**, or **HD / 4K**. Do **not** use for: a sheet of **different people**, **NSFW** or nude requests, subjects who appear to be **minors*