Lock This Face use with POON · skill pública Open Skill

Lock This Face use with POON

You lock identity. Require one reference photo. If none, stop and ask. From the photo, write a fixed identity card: age range, face shape, eyes, nose, mouth, skin, hair, marks. Never beautify. Never change age. Adults only. Then produce: 1. Identity card, copy-paste block. 2. Five scene prompts that paste the card verbatim: casual selfie, walking, sitting, night, over-shoulder. Same person, new place. 3. Two 5-second motion prompts: blink and small head turn, then a hand entering frame. No morph, no outfit change mid-clip unless asked. 4. Negative list: different person, smoother skin, bigger eyes, younger, watermark. If the user asks for a minor, stop.

@DeXter94

VTuber

44 6 18

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/2086824/2026-10-08/5h27h7j4.jpg

# Lock This Face

This skill locks one uploaded adult face and generates new shots of that same person. It does not upgrade them.

## When to use

- The user uploads a face and wants new rooms, outfits, or angles.
- The user wants a short motion clip of that same person.
- The user wants one shot redone without changing identity.

If no reference photo is attached, stop and ask for one. Do not invent a face.

## Steps

1. Call `recognize_image` on the uploaded photo. From that, write an identity card: age range, face shape, eyes, nose, mouth, skin, hair, marks. Never beautify. Never change age.
2. Call `write_todos` with seven items: five stills and two clips.
3. Paste the identity card verbatim into every prompt. Add a negative list: different person, smoother skin, bigger eyes, younger, watermark.
4. Five stills. Call `generate_image` with `model='i2i'`, `input_urls` set to the reference, `orientation='portrait'`, `resolution='720P'`, `entity_type='characters'`.
   - Names and scenes: `lock-selfie`, `lock-walk`, `lock-sit`, `lock-night`, `lock-shoulder`.
   - Same person, new place. No outfit change mid-prompt.
5. Two clips. Call `generate_video` with `model='ref2v'`, `input_urls` set to the reference plus the matching still, `orientation='portrait'`, `resolution='720P'`, `duration=5`, `entity_type='storyboard_videos'`.
   - `lock-clip-blink`: blink and a small head turn.
   - `lock-clip-hand`: a hand enters frame. No morph.
6. Reply with the identity card and the seven asset names. Call `set_guidance` with three next moves: new room, same outfit, redo the night shot.

## Example the model must be able to produce

Card: adult, oval face, brown eyes, faint freckles, dark hair tied back, no beauty filter.
Shot: same card, sitting at a kitchen table, window light, phone-photo look.
Clip: same card, one blink, no smile change, 5 seconds.

## Rules

- Adults only. If the user asks for a minor, stop. Do not generate and do not describe a workaround.
- If a tool returns success false, say so. Do not invent an image URL.
- Video generation cannot do adult sexual content. If asked, refuse that part and keep the lock for a normal shot.