Memorial Yayoi Kusama

把你的照片做成“一颗圆点，无限的爱”风格的原创纪念图。

Reimagine the pictured subject as a figure, creature, object, or environment
within a clearly recognizable Yayoi Kusama visual universe. The source photo
supplies the emotional subject and distinctive cues; it does not have to remain
a literal photographic portrait. The goal is not to preserve every pixel, but
to preserve who or what is being remembered while transforming its form,
material, scale, and surroundings through Kusama's broader visual language.

This is a visual transformation workflow, not a generic image prompt. Use an
image-understanding capability to inspect the source and an image-generation or
image-editing capability to produce candidates. Keep the original file intact.
The bundled assets/style-reference.png is the default style benchmark and
must travel with the skill.

Workflow

1. Accept exactly one source photo. It may contain a person, pet, object,
   building, landscape, plant, artwork, or a meaningful place. Do not require a
   portrait.
2. Ask only one optional question if the user has not supplied it:
   你想把哪个词送向未来？
   Accept one word or a short phrase, and allow the user to skip it. Do not
   invent a eulogy, poem, quotation, date, signature, or message.
3. Analyze the image before generating. Identify the subject's emotional and
   distinctive cues: recognizable facial or animal features, silhouette, pose,
   gesture, markings, color relationships, objects, setting, and the feeling of
   the moment. Preserve recognizability while allowing genuine artistic
   re-formation. Do not equate broader Kusama vocabulary with adding many
   decorative motifs. Select one dominant mark-making system whose repetition
   and accumulation transform the subject and the whole picture plane.
4. Build an image prompt using the rules in references/visual-rules.md.
   Use assets/style-reference.png as a separate style reference in every
   generation. The source photo controls subject identity and composition; the
   bundled benchmark controls only medium, line, flatness, mark rhythm, palette,
   and spatial construction. Transfer those abstract traits, not its depicted
   figures, words, watermark, or layout. If the image tool accepts multiple
   references, pass the source first and the bundled benchmark second, and label
   their roles explicitly in the prompt. If it accepts only one image, pass the
   source and translate the benchmark traits into text using
   references/visual-rules.md. Name the visual language directly as a
   recognizable Yayoi Kusama visual grammar while requiring an original
   composition. Explicitly suppress polished CGI, anime beauty rendering,
   fantasy concept art, and children's-book illustration. Do not use any
   specific artwork, installation, exhibition, or image as a composition target.
5. Generate 2 or 3 internal candidates. Keep the source photo as a reference
   in every candidate. Prefer preserving the source aspect ratio.
6. Inspect and score every candidate with references/quality-evaluation.md.
   Reject any hard failure. Return only the highest-scoring passing candidate;
   do not expose the other candidates unless asked.
7. If the best candidate has one correctable weak dimension, perform at most
   one targeted revision. Phrase the revision as a constraint, such as
   preserve the subject's distinctive face cues and gesture; reduce decorative dots around the eyes; replace the literal background with a biomorphic field while keeping the memorial mood. Re-score the revised result.
   If the subject has drifted, return to the original source rather than
   repeatedly redrawing it.
8. If no candidate passes, say that the memorial image could not be generated
   reliably and explain the failed dimension briefly. Do not deliver a visibly
   distorted result.

User Adjustments

After returning a passing result, accept constrained requests such as:

• make the figure flatter, rougher, more hand-painted, or more plainly drawn;
• reduce motif variety and commit to one dominant system of irregular marks,
  dots, nets, or repeated contours;
• preserve more of the source subject's face, markings, pose, or silhouette;
• change the subject from a literal portrait into a more transformed Kusama-like
  character while keeping it recognizable;
• make the subject more visually dominant or the surrounding infinity more
  expansive.

Treat 再来一版 as a request to revise the current result only when the user
also gives a direction. Otherwise ask what should change. Every revision must
retain the source subject's distinctive cues, but it may continue the artistic
re-formation. Keep the prior passing result available as the fallback.

Output

Return:

1. The final image as a local artifact, without overwriting the source.

2. A short explanation outside the image describing the memory -> one dot -> infinity transformation and mentioning the supplied word only if one was
   supplied.

3. This independent-project statement:

   这是一个独立的参与式纪念项目，使用具有明显草间弥生辨识度的原创视觉语法，受到重复、无限、爱与希望等主题启发。它不复制任何具体原作，也不代表草间弥生本人、草间弥生美术馆、草间弥生基金会或草间弥生工作室。

4. The official museum notice:
   https://yayoikusamamuseum.jp/en/news/news20260827/

5. One concise, source-grounded fact about Kusama's wider practice beyond
   dots, selected from painting, sculpture, installation, performance, film,
   fiction, or poetry. Verify current factual claims from an official or
   reputable cultural source before stating them.

Keep the source image, prompt, output, and optional word local to this run by
default. Do not upload them to a public gallery, retain EXIF location data, or
claim official collaboration or authorization. If image metadata can be
removed, remove it from the delivered derivative; never alter the source.

Boundaries

The result should be unmistakably legible as Kusama-inspired in its visual
vocabulary, not merely as a dot overlay or decorative fantasy illustration.
Use a controlled subset of her recurring concerns: obsessive repetition and
accumulation, hand-varied marks, nets, altered scale, self-obliteration, and the
tension between a singular subject and a larger field. Prefer flat matte paint,
visible human irregularity, blunt black contours, compressed picture space, a
restricted saturated palette, and one repeated unit that crosses subject and
background. Motif proliferation, polished CGI highlights, smooth plastic
modeling, anime beauty rendering, cute storybook imagery, and generic cosmic
scenery weaken the result even when they contain dots. Do not reproduce a
particular Kusama artwork or installation, including a pumpkin, infinity mirror
room, recognizable sculpture, exhibition layout, or signature. Do not generate Kusama's face, voice, handwriting,
quotation, or invented final words. Do not add logos, watermarks, dates,
commercial copy, or promotional language.

Read references/visual-rules.md when composing or revising a prompt, and use
assets/style-reference.png as the default style benchmark. Read
references/quality-evaluation.md when selecting candidates. Read
references/rights-and-sources.md when handling source links, public sharing,
rights questions, or factual biography.