Skip to content

1173206772/photo-skills / skills/genshin-inspired-photo

genshin-inspired-photo

Verified

Transform supplied photos or written concepts into luminous Genshin Impact-inspired anime-fantasy imagery while preserving the original subject, identity, pose, composition, architecture, landscape, or food structure. For people, default to dimensional 3D anime game rendering and function-preserving redesign of modern clothing and accessories so the person belongs in the selected nation without becoming a copied character. Includes nation-specific direction for Mondstadt/蒙德, Liyue/璃月, Inazuma/稻妻, Sumeru/须弥, Fontaine/枫丹, Natlan/纳塔, and the northern Nod-Krai/Snezhnaya direction. Use for people, portraits, outfits, pets, landscapes, buildings, streets, interiors, vehicles, objects, and dishes when the user asks for 原神风, 原神3D人物, a named Genshin nation, Teyvat-like fantasy, elemental anime game art, cel-shaded fantasy photography, or a bright open-world anime RPG look.

references/prompt-recipes.md

View on GitHub
# Prompt Recipes

## Iterative Lineage Lock

For follow-up edits, do not treat the newest output as the universal source of truth. First identify the authoritative image for each layer:

```text
Layer provenance:
- Composition and camera: IMAGE [n]
- Hero identity, pose and anatomy: IMAGE [n]
- Approved clothing and equipment: IMAGE [n]
- Approved environment and architecture: IMAGE [n]
- Requested change: [one layer only]
```

If the approved background and approved subject come from different images, use a compositing edit. Explicitly say which image is the edit target, which is a supporting appearance source, what must remain unchanged, and what must be transferred. Never preserve a known regression merely because it appears in the latest image.

## Edit Template

```text
Use case: style-transfer image edit.
Primary request: Transform the supplied photo into an original Genshin-inspired luminous anime-fantasy scene.
Input image: The supplied image is the edit target and composition reference.
Layer provenance: [name the authoritative image for composition, hero, approved clothing/equipment, environment and architecture]. Change only [requested layer].
Hero subject: [person / animal / landscape / building / interior / food / object] performing or showing [source action or state].
Preserve exactly: [identity, expression, pose, hands, clothing silhouette, object count, architecture, terrain, ingredients, camera viewpoint, crop and aspect ratio].
Transformation strength: [Faithful paint-over / World conversion / Key-art cinematic].
Landscape anchor strength: [Subtle only when requested / Readable by default / Strong when recognition is weak]. For Readable or Strong, specify one primary midground landmark, its visual scale, a connected exploration route and one functional elemental system.
Composition: Preserve [source framing and geometry]; maintain clear foreground, midground and distance where present.
Nation direction: [Mondstadt / Liyue / Inazuma / Sumeru rainforest / Sumeru desert / Fontaine / Natlan / Nod-Krai or source-derived compatible direction]. Express it through [terrain], [architecture or object construction], [materials], and [clothing or food presentation].
Subregion variant: [none or one specific variant].
Anchor stack: include at least four mutually reinforcing families from [3D anime renderer], [nation spatial/material grammar], [elemental world function], [character/civilian construction], and [exploration composition]. Do not count color or particles alone.
Functional translations: [source item] remains [protected function, position, volume and interaction] but changes [two or more construction systems] into [nation-compatible constructed equivalent]. Remove branding only from [items]. Convert every visible background person into a simplified nation-compatible 3D anime civilian.
Rendering mode: [3D anime world render by default when a person is present / 3D environment render / illustrated key art only when requested].
Rendering: coherent modeled depth across subject and environment, sculpted anime forms, grouped modeled hair, garment thickness and seams, hand-painted albedo textures, ambient occlusion, two or three controlled toon-shadow regions, selective material highlights, contact shadows and environmental bounce light.
Elemental direction: [one primary element], expressed through [motivated physical paths].
Lighting and palette: [specific natural light and elemental palette], with reflected effect light on nearby materials.
Constraints: Original fan-style design; keep anatomy, materials and physical structure coherent; no unrelated fantasy props.
Avoid: copied official characters, costumes, weapons, companions, city layouts, logos, UI, icons, text, watermark, UID, generic chibi, thick black outlines, plastic skin, excessive bloom and malformed details.
```

## Follow-Up Edit and Multi-Image Repair Template

```text
Use case: compositing / identity-preserve correction.

Image roles:
- IMAGE 1 is the edit target and supplies [approved composition/background/architecture].
- IMAGE 2 is the appearance source and supplies [approved hero/clothing/equipment/object]. Do not copy IMAGE 2's background.

Primary request: Keep IMAGE 1's approved [layers] exactly. Replace or restore only [requested layer] using IMAGE 2. This corrects a regression; do not redesign unrelated content.

Background lock — keep from IMAGE 1:
- Preserve [terrain, architecture, route, lighting, weather, crop and camera] as closely as possible.
- Do not remove, relocate, repaint, simplify or restyle these approved elements.

Subject transfer — take from IMAGE 2:
- Transfer the complete approved [garment/equipment/object] construction, materials, attachments, volume, key colors and renderer.
- Place it on the subject at IMAGE 1's exact identity, pose, anatomy, scale and position.
- Name the distinctive approved construction details that must return; do not summarize them only as “fantasy clothes” or “game camera.”

Integration: Change outside the transferred silhouette only for minimal contact shadow, reflected light, atmosphere and edge cleanup. Match IMAGE 1's key light and perspective.

Final lock: IMAGE 1 background and composition remain authoritative. IMAGE 2 subject appearance remains authoritative. No regression to modern clothing, accessories or equipment; no copied character, logo, UI, text or watermark.
```

For a single-image follow-up, use the same structure but omit the transfer block. A background-only request must explicitly lock identity, pose, approved clothing, equipment, accessories and their construction—not merely say “keep the person.”

## Generation Template

```text
Use case: stylized-concept image generation.
Primary request: Create an original luminous anime-fantasy scene inspired by the visual language of Genshin Impact.
Hero subject: [describe subject and action].
Setting: [landscape, architecture, interior or table setting].
Composition: [camera, crop, focal position, depth layers and exploration path].
Elemental direction: [one primary element and physical motif path].
Nation direction: [one nation and optional subregion], expressed through coherent landform, architecture, materials, clothing, food presentation and atmosphere—not a token prop.
Rendering mode: [3D anime world render / 3D environment render / illustrated key art when explicitly requested].
Rendering: readable anime shape design, modeled dimensional lighting, controlled toon shading, hand-painted albedo materials, ambient occlusion, material-specific highlights and atmospheric open-world depth.
Palette and mood: [colors and emotion].
Constraints: Entirely original people, costume, architecture, food, symbols and props; natural anatomy and physics; no branding or text.
Avoid: direct official character or landmark reproduction, logo, UI, rarity stars, UID, watermark, over-saturation, generic fantasy clutter and flat poster background.
```

## Person Recipe

```text
Keep the person's real identity, age, facial proportions, expression, hairstyle, pose, hands, role and broad clothing silhouette unmistakable. Render the entire person as a dimensional 3D anime game model without enlarging the eyes excessively: stable sculpted facial planes, modeled grouped hair, layered garment geometry, seams and closures, material-specific highlights, ambient occlusion and contact shadow. Preserve key clothing colors. Translate conspicuously modern garments and accessories into functionally equivalent nation-compatible designs while keeping position, carried volume, straps, grip and interaction. Keep the background connected through matching key light, rim light, bounce color and atmospheric depth. Do not replace the person with an existing game character.

For each modern item, name at least two construction changes. Do not accept an unchanged sportswear, baseball-cap or DSLR silhouette with fantasy trim. Convert every visible background person too; distant people may be simple, but their clothing silhouettes and materials must belong to the selected nation.
```

## Landscape Recipe

```text
Preserve the exact terrain silhouette, horizon, camera and existing travel path. Set anchor strength explicitly. In the default Readable mode, build a foreground route cue, one clear source-compatible midground landmark or functional structure, and a distant promise. Make the route lead toward or through the landmark. Make one elemental system visibly affect the route, structure, terrain, vegetation, water or weather. Repaint vegetation as clustered hand-painted masses and strengthen large cloud and mist layers without hiding the anchor. Keep the scene original, explorable and coherent. No copied game landmark, generic fantasy castle, floating UI, token prop, particle-only solution or random character insertion.
```

## Strong Landscape Recognition Correction

Use this block when feedback includes “太克制,” “看不出来是原神,” “锚点不够,” “更像游戏,” or equivalent:

```text
Correction target: Increase Genshin-inspired open-world recognition without changing the source camera, horizon, terrain silhouette, road or river course, major vegetation masses, weather logic or approved subjects.
Anchor strength: Strong. Place one original nation-specific landmark or functional world structure at readable midground scale—roughly 10–25% of frame width or comparable visual weight—with a clean silhouette and source-compatible placement. It must not be confined to a tiny lower corner or hidden by fog.
Exploration chain: Create one visible foreground cue that leads toward or through the primary anchor, then continues toward a distant destination or atmospheric promise.
Elemental function: Show [element] doing useful physical work through the route, structure or terrain, with visible cause and effect and local reflected light. Do not satisfy this requirement with loose particles alone.
Nation construction: Express [nation] through at least three large-scale systems such as landform adaptation, structure silhouette, materials, public utility, settlement rhythm or vegetation management. Do not rely on one color, one windmill, one gear, one gate or one crystal.
Thumbnail gate: At roughly 20% display size, the primary anchor, route, nation grammar and elemental function must still be readable. If not, enlarge or simplify the anchor rather than adding clutter.
Originality and preservation: Keep all designs original; do not copy an official landmark, character, costume, icon, UI or logo. Preserve approved clothing, equipment, identity, pose and background lineage.
```

## Architecture Recipe

```text
Preserve the building footprint, floor count, roofline, windows, doors, structural supports and camera perspective. Amplify its existing cultural cues with one compatible regional fantasy family. Use ornament at focal edges, functional magic or machinery, clean painted materials and atmospheric depth. Do not replace the building type or turn every surface into decoration.
```

## Food Recipe

```text
Preserve every recognizable ingredient, portion, doneness, container and viewing angle. Repaint the dish with clean anime-game shapes, appetizing painterly volume, warm bounce, moist glaze or crisp texture, and a restrained ingredient-derived elemental garnish. Keep it physically edible. No extra ingredient, plastic toy look, floating UI, stars, labels or fake text.
```

## Personalization Mapping

- **“更像照片 / more photographic”** — retain realistic proportions and lens depth; reduce outline, costume changes and particles.
- **“更像游戏 / more game-like”** — strengthen shape grouping, cel-shadow boundaries, painted materials and elemental atmosphere while preserving identity.
- **“太克制 / 看不出来是原神 / 锚点不够”** — switch landscape anchor strength to Strong; enlarge and clarify one nation-specific midground structure, connect it to a visible route and functional elemental system, and apply the thumbnail gate. Do not answer with more particles.
- **“保留衣服 / keep the clothes”** — change only rendering and light; do not add fantasy garment parts.
- **“更像原神人物 / more like an in-game character”** — use the full 3D anime world render specification, stronger garment construction and functional nation translation; do not copy a named character.
- **“改服装但功能不变 / redesign clothes, preserve function”** — protect role, movement, weather protection, storage, viewing or recording function, attachment and key color; redesign only silhouette details, panels, material and ornament.
- **“不要魔法 / no magic”** — use regional architecture, color and painterly rendering without particles or elemental glow.
- **“更华丽 / more ornate”** — increase focal embroidery, jewelry, railing, roof-edge or plating detail; keep quiet surfaces elsewhere.
- **“更可爱 / cuter”** — soften shapes and palette modestly; do not force chibi proportions unless explicitly requested.
- **“更史诗 / more epic”** — widen depth, strengthen landmark scale, wind and directional light; do not change the hero pose or identity.
- **“完全原创 / fully original”** — avoid all named characters, recognizable costumes, landmarks, symbols and region-specific copies; use only abstract visual grammar.
- **“蒙德 / Mondstadt”** — route to the Mondstadt card in the nation atlas; do not merely add windmills.
- **“璃月 / Liyue”** — route to the Liyue card; preserve Chinese spatial depth and material logic without mixing in Inazuma shrine cues.
- **“稻妻 / Inazuma”** — route to the Inazuma card; choose one island mood instead of making every surface purple.
- **“须弥 / Sumeru”** — choose rainforest, desert, or a deliberate transition between them.
- **“枫丹 / Fontaine”** — distinguish surface metropolis, rural waterscape, underwater world, and industrial interior.
- **“纳塔 / Natlan”** — choose terrain and community logic first; avoid a generic “tribal” costume collage.
- **“至冬 / 诺德卡莱 / Snezhnaya / Nod-Krai”** — use the cautious northern card and do not invent unreleased national landmarks or royal symbols.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 196kUpdated Claude CodeCodexVerifiedInstall