Dark Subculture Restyle · skill pubblica Open Skill
Dark Subculture Restyle
Dark Subculture Restyle è una Open Skill pubblica su Digen. Aprila per vedere cosa crea questo workflow.
Story
0 0 1
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4150474/2026-10-15/px3e5dbr.jpg
# Dark Subculture Restyle
Restyle the image the user gives you into a **dark subculture** anime look: underexposed, cold, desaturated, with crimson accents — the subject, pose and composition of the original are preserved, only the rendering style changes. One reference image in, one image out.
Write the image `prompt` in **English**; talk to the user in their own language.
## When to use
- The user attaches an image and asks to restyle / redraw / convert it into a dark, alt, goth, subculture, moody or "dark anime" look.
- The user asks for a *new* image in this style and gives no reference image (use the text-to-image branch in step 3).
- The user asks for another pass on a restyle from earlier in this conversation ("darker", "keep the face softer", "try the sharper variant").
## Input and output
- **Input**: one reference image the user uploaded, or an image asset already generated earlier in this conversation. If several were attached, use the most recent one the user is clearly pointing at; if it is genuinely ambiguous, ask which one.
- **Output**: one image, same orientation and composition as the input. It is added to the canvas automatically — never paste image URLs back to the user.
## Step 1 — Read the input image
1. Identify the reference image reference (`asset_id` / `s3://` uri / https URL exactly as it appears in the conversation). **Never invent or reconstruct a URL.**
2. When the image content is unclear or you need to lock the layout down, call
`recognize_image(image_url=<reference>, prompt="Describe the subject, pose, framing/crop, clothing, background and lighting in short bullet points.")`
and reuse its wording inside the *preserve* clause below.
3. If there is no image and the user did not ask for a from-scratch generation, ask them to upload one.
## Step 2 — Build the English prompt
Always use this five-part shape, **in this order**: *preserve → style → character → background → finish → avoid*. Keep it roughly 100–200 words; if it gets long, cut the character/background detail, but **never** drop the preserve clause, the low-key lighting clause, or the crimson-accent palette clause.
Fill the `<...>` slots from the user's image and request:
```
Dark subculture restyle of the reference image. <subject / pose / framing preserved from the input, e.g. "the same elf-eared girl with twin silver ponytails standing between two dark wooden doors, hands on the door handles, half-body vertical framing, white cape with faded gold trim, black-and-white striped top, black belt, red gem jewelry, matching the input exactly">. Keep everything that identifies the input: same composition, crop, camera angle, subject count, pose, gesture, silhouette, body proportions, outfit design and colour blocking, background structure and props — nothing added, removed or replaced.
Style: dark subculture / dark alt-goth anime illustration, low-key and underexposed, crushed deep blacks, heavy chiaroscuro, one soft rim light so illumination stays only where the viewer must read (face, hands, fabric edges), large areas of shadow with almost no detail.
Palette: cold and desaturated — ash grey, bone white, blue-grey; crimson / oxblood accents only on eyes, jewellery and trim; muted antique gold; no bright or warm saturated colours.
Character: porcelain pale desaturated skin, sunken shadowed eye sockets, dark under-eye shading with smudged red eyeshadow, heavy dark eyeliner, tired hollow gaze with one small red iris highlight, muted lips; silver-white or black hair with loose flyaways; layered alt fashion — long cape or oversized coat with faded gold trim, black-and-white striped top, choker, leather belt with metal buckle, straps, gem pendants, lace, dark thigh-highs.
Background: keep the original environment, pushed darker, cooler and slightly blurred — dim stone doorway, brick, trees or interior as in the input, only faintly readable behind the subject.
Finish: anime semi-realistic painterly rendering, fine clean linea