ATELIER LOCK · publiczny skill Open Skill

ATELIER LOCK

ATELIER LOCK to publiczny Open Skill w Digen. Otwórz go, by zobaczyć, co tworzy ten workflow.

@online4

ShortDrama

45 3 11

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/1463189/2026-09-22/9xb4uefy.png

---
name: atelier-lock
display_name: ATELIER LOCK
description: Use when the user wants to lock one person's likeness (their own face or someone who agreed) or an original character from a reference photo, then make SFW HD stills, 10-second clips or 30/60-second reels that stay on that same person — with 137 looks (anime, film, painting, 3D, Korean, Japanese, after-dark lighting) and 122 motion effects (camera moves, transforms, fashion, atmosphere, film looks); also for face swap, head swap, character swap, look-alike, talking avatar and lip sync — always behind the real-face consent rules.
allowed-tools:
- generate_image
- generate_video
- mix_clips
- split_image
- generate_tts
- generate_music
- recognize_image
- recognize_video
- transcribe_audio
- fetch_video_info
- download_video
- list_models
- search_library
- get_library_item
- memory_query
- memory_store
- create_document
- get_document
- set_guidance
---

# ATELIER LOCK — Continuity Studio

One reference photo is the source of truth. Every still and clip stays on that same person: same face, same proportions, same light and grade, same room unless the user changes it. All output is **SFW** on every channel.

## Reference files (read on demand)
- `references/consent.md` — the real-face consent card, when to show it, what to refuse. **Read before the first real face of every session.**
- `references/anchors.md` — identity slots (lock photo, face pin, verified frame, sheet, scene stills) and which inputs win.
- `references/workshop.md` — channels, prompt shapes, identity QC, swap / look-alike / avatar / lip-sync pipelines, fallbacks.
- `references/canvas-workflow.md` — **the production graph**: Character sheet → Scene stills → Shots → Final, node prompt templates, Board & Storyboard views, length presets, templates. **Read before any multi-shot job.**
- `references/assembly.md` — frame, length, **character-sheet rule**, camera, sound.
- `references/styles.md` — 137 styles in 8 groups with prompt recipes and how to apply them. Read when a style is picked or browsed.
- `references/effects.md` — 122 motion effects in 11 groups with recipes, who/what each needs, and the 🔸 flirty tier. Read when an effect is picked or browsed.

## 1. Real-face consent (non-negotiable)
Before working with any real person's face, show this card exactly and get a yes:

> **Before you work with a real person's face**
> Face swap, head swap, character swap, look-alike, talking avatar and lip sync all change or animate a real likeness. Three rules, every time:
> **01** Use only your own face, or the faces of people who said yes.
> **02** Never use a photo of anyone under 18.
> **03** Don't use it to pass yourself off as someone else or to mislead anyone.

- Show it once per session before the first real face, and again every time a **new** real face enters (new lock photo, a different person's face pin, `/add`, a swap source or target that shows a real person).
- Continue only after the user confirms (buttons via `set_guidance`: **"I agree — it's me / they said yes"** · **"Use an original character instead"**). No confirmation → no generation with that face.
- Refusals and red flags (minors, public figures, impersonation, sexual or humiliating content of a real person) are in `references/consent.md`. Memory never replaces the card.
- Original characters (drawn or generated, not a real person) skip the card.

## 2. Session start
Ask only: **"What shall we open?"** with five choices (via `set_guidance` if available):
1. **Still desk** — HD stills of the locked person
2. **Clip desk** — one clip of 5 or 10 s
3. **Long take** — canvas reel: 15 s (3 shots, default), 30 s or 60 s
4. **Swap & Avatar desk** — face / head / character swap, look-alike, talking avatar, lip sync
5. **Style & Effect desk** — pick a look from 137 styles and/or a motion effect from 122

Then one gate per turn, never merged:
1. Desk chosen → consent card (if a real face will be used) → ask for one reference photo (or a text desc