Enter Spooky Town · সর্বজনীন Open Skill
Enter Spooky Town
Enter Spooky Town Digen-এ একটি সর্বজনীন Open Skill। খুলে দেখুন এই ওয়ার্কফ্লো কী তৈরি করে।
Roleplay
2 0 2
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4284865/2026-10-21/ncwy5x2b.jpg
# Enter Spooky Town Turns one uploaded portrait into a dark, cinematic Halloween still: the person keeps their face, hairstyle, expression and pose, but is re-dressed in gothic Halloween attire and dropped into a foggy moonlit spooky town. There is no project state — work from the image and text in this conversation, and put the result on the canvas. ## When to use - The user uploads a portrait (selfie, headshot, full-body photo) and wants a Halloween / spooky / gothic makeover of it. - The user asks to change the outfit and background of an uploaded photo into a Halloween town scene. - The user asks to redo or adjust a previously generated Spooky Town result (more vampires, different costume, different town mood). If the user has no image at all, see `references/step-details.md` (Branch B) before generating. ## Inputs and outputs - Inputs: one reference photo (the person to keep) plus optional free text describing costume, mood, or extra props. - Output: one image, `entity_type='characters'`, with a short `name` (e.g. `Spooky Town Portrait`). ## Steps 1. Identify the reference image in the conversation. It is the source of identity — it must be passed through `input_urls`, never described in words instead of being passed. 2. Read `references/style-and-prompt.md` with `read_skill_file` and follow its Style / Subject / Layout / Must keep / Must avoid rules and its prompt template. 3. Read `references/step-details.md` with `read_skill_file` for the input checks, the orientation and resolution decision, the prompt assembly order, and the retry / failure branches. 4. Write one English prompt by filling the template from `references/style-and-prompt.md`: keep clauses for identity first, then costume, then the town layout and lighting, then the cinematic grade. Fold any user text in as additive modifications without weakening the must-keep clauses. 5. Call `generate_image` with `prompt=<the assembled English prompt>`, `input_urls=[<reference image>]`, `model='i2i.hd.lite'`, `orientation` per the decision table, `name='Spooky Town Portrait'`, `entity_type='characters'`. Do not pass `width`/`height`, and do not pass a physical model name. 6. Reply in natural language: what was kept (face, hair, pose), what was changed (costume, town, lighting). Do not paste long URLs. ## Notes - Identity is the hard constraint. If the result no longer looks like the uploaded person, regenerate once with the identity clause moved to the front of the prompt and strengthened; never claim success on a mismatched face. - Video generation is out of scope for this skill; only `generate_image` is used. - `i2i.hd.lite` does not support NSFW content. Decline adult, gory or gore-horror requests politely and do not call the tool. - If `success=false`, report the real error; never invent an image URL or an asset id. - Do not add elements the app does not have (no text overlays, no title lettering, no collage, no multi-person group shots).