AI Japanese Christmas — Tokyo Holiday Sticker Collage · सार्वजनिक Open Skill

AI Japanese Christmas — Tokyo Holiday Sticker Collage

AI Japanese Christmas — Tokyo Holiday Sticker Collage Digen पर एक सार्वजनिक Open Skill है। खोलें और जानें कि यह वर्कफ़्लो क्या बनाता है।

@eggs

Doc

0 0 0

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4284862/2026-10-16/nk5a8g43.jpg

# AI Japanese Christmas — Tokyo Holiday Sticker Collage

Turns one reference photo of a person into one finished collage image: the same person repeated as die-cut photo stickers on a kawaii Christmas scrapbook page, with Tokyo winter scenes, holiday doodles and short Japanese / English text stickers.

This is an **image-to-image** skill. There is no structured project state; the collage is the deliverable.

## When to use

- The user uploads or points to a photo and asks for a Japanese Christmas / Tokyo Christmas / holiday sticker collage, sticker sheet, kawaii scrapbook or "sticker version" of it.
- The user asks to redo, restyle, recolor, or make a different aspect-ratio version of a collage already produced with this skill.
- Do **not** use it for a plain portrait, a normal single Christmas photo, or a text-only request.

## Input and output

- **Input:** one or more photos of a person (selfie / portrait). Identity is the only thing that must survive exactly: same face, eyes, eyebrows, hair colour and hairstyle, skin tone, body type.
- **Output:** one image — a dense vertical sticker-sheet collage. The reference photo is used for identity and styling only; its own background is discarded and replaced by the scrapbook page described in the spec.

## Steps

1. Confirm the request and locate the reference photo(s). If the user gave a text-only request with no photo, ask for a photo — do not switch to text-to-image.
2. Read the visual law before writing any prompt:
   `read_skill_file(name='tokyo-christmas-sticker-collage', path='references/style-and-prompt.md')`
   It holds Style, Subject/content, Layout, the doodle and text-sticker sets, Must keep / Must avoid, and the copy-paste prompt template.
3. Read the operational details:
   `read_skill_file(name='tokyo-christmas-sticker-collage', path='references/step-details.md')`
   It holds the input-analysis checklist, prompt assembly order, retry rules and the delivery checklist.
4. Analyse the reference photo. Look at it directly; call `recognize_image` once only if it arrives as a URL / asset id that you cannot see, or if you need a closer read of identity details. Write down 3–6 identity anchors (face shape, eye shape, eyebrows, hair colour / length / fringe, skin tone, makeup) to repeat verbatim in the prompt.
5. Assemble the English prompt from the template in `references/style-and-prompt.md`, filling its placeholders: identity anchors, sticker count and grid, scene list, doodle set, text-sticker phrases, palette. Follow the assembly order in `step-details.md`.
6. Generate the collage:

   ```
   generate_image(
     prompt   = <assembled prompt>,
     input_urls = [<the reference photo: asset_id / s3:// uri / https>],
     model    = 'i2i.hd.lite',
     orientation = 'portrait',
     resolution  = '1080P',
     name     = 'Tokyo Christmas Sticker Collage',
     entity_type = 'sticker_collages'
   )
   ```

   Default is a vertical sheet (`orientation='portrait'`, `resolution='1080P'`). Switch to `square` or `landscape` only when the user asks for it; keep the same `model` channel. Never pass a physical model name.
7. Check the returned image against the delivery checklist in `references/step-details.md`. If identity or the sticker-sheet look failed, apply the matching fix from that file and regenerate — at most two retries. If it still fails, deliver the best result and say honestly what is off.
8. Reply with 1–3 natural sentences about the collage (scenes used, identity preserved, festive mood). Do not paste image URLs or long prompts. If useful, call `set_guidance` in the same turn as the final reply with 2–3 concrete follow-ups (e.g. square version, red-and-green palette, more night-illumination shots).

## Notes

- Never change the person's identity, and never add other people: every sticker shows the same person as in the reference photo.
- Do not invent content that is not in the spec — no brand logos, no extra cities, no scenes outside the sticker list.
- Refuse NSF