Anime Illustration — Open Skill
Anime Illustration
Anime Illustration is a public Open Skill on Digen. Open it to see what this workflow creates.
Custom
0 0 0
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4150474/2026-10-15/gbmh5hcm.jpg
# Anime Illustration Turn a text description into a high-fidelity anime illustration with clean lineart, soft cel-shading, unified color design and a cozy slice-of-life mood. There is no project state: every request produces one finished image. ## When to use - The user describes a character, a scene or a mood and wants it drawn as anime art. - The user asks for an avatar, portrait, key visual, poster or wallpaper in anime style. - The user wants a previously generated anime illustration re-rendered with a different pose, outfit, setting or framing. ## Steps 1. Read the request and settle four things before writing anything: **subject** (who or what is in frame), **setting** (where), **mood** (tone/lighting) and **framing** (bust / full body / wide scene). Ask at most one short clarifying question if the subject itself is ambiguous; otherwise proceed directly. 2. Write the image prompt **in English**, always in this order: `<subject + key features> — <pose / action> — <setting> — <lighting + mood> — <style tags>` Keep this style-tag block in every prompt — it is the visual identity of this Skill: `anime illustration, clean lineart, consistent facial features, hairstyle and outfit design, soft cel-shading, unified color palette, detailed background, cozy slice-of-life mood, high fidelity` Prompt rules: - **Subject**: only the subject(s) the user described. Do not invent extra characters, mascots, logos or on-image text. - **Character consistency**: when the user asks for more variants of the same character, repeat the *identical* wording for face, hair and clothing in every prompt. Never re-describe the same character differently between generations. - **Layout**: state the composition explicitly, e.g. `centered bust shot, three-quarter view`, `full-body standing pose on the left third, layered foreground and background`, `single focal subject with generous negative space`. Append `no text, no watermark, no signature` unless the user explicitly asks for a poster title. - **Content policy**: if the request is sexual, violent, hateful or otherwise disallowed, politely decline and do not call any tool. 3. Call `generate_image` with: - `prompt`: the English prompt built in step 2 - `model='t2i'` — the text-to-image channel. Use `model='t2i.hd'` only when the user explicitly asks for extra fine detail; note that `.hd*` channels do not accept NSFW content. - Reference-image case: if the user supplied an image, keep it in `input_urls` and switch the channel to `model='i2i'` (or `model='i2i.hd'`) instead of `t2i`. - `name`: a short 2–5 word title (character name or scene title) so the canvas card is readable. - `entity_type='illustrations'` - `orientation`: default `portrait` for character art and standing figures; `landscape` for wide scenes and key visuals; `square` for avatars and icons. - `resolution`: omit it unless the user asks for a specific quality level; the platform default (`720P`) applies. - Use plain text descriptions only; do not pass physical model names or pixel dimensions. 4. Reply with one or two natural-language sentences: what was drawn, which style choices were applied, and one concrete next variation (a different pose, outfit or setting). Do not paste long URLs — the illustration is already on the canvas. ## Notes - If `generate_image` returns `success=false`, tell the user honestly and quote the reason. Never fabricate an image URL or claim an image exists. - One prompt produces one illustration; do not promise a batch or repeat the same call to "get more". - Every prompt stays in English. Keep the user's own proper nouns (character names, place names) untouched inside the prompt. - Generated assets land on the canvas automatically — describe them in words, do not dump links. - Video generation, voice-over, music and video editing are out of scope for this Skill.