Jack-o'-Lantern · Open Skill 公開スキル
Jack-o'-Lantern
Jack-o'-Lantern は Digen 上の公開 Open Skill です。開くと、このワークフローが何を作るか分かります。
Manhua
1 0 0
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4150474/2026-10-15/jawwntbb.jpg
# Jack-o'-Lantern Turn one reference image into a single photorealistic Halloween photograph: a real hand-carved pumpkin whose carving is a motif of the reference character, met by that character on a roadside at night, the way you would photograph a genuinely made "real jack-o'-lantern". There is no structured state here — the deliverable is one image asset plus a short natural-language reply. ## When to use - The user attaches an image and asks for a jack-o'-lantern, carved pumpkin, Halloween lantern, or "Halloween version of this photo". - The user wants a fictional character, avatar, OC, pet or mascot reimagined as a carved pumpkin. - The user wants to restyle / re-render a lantern image that already exists in this conversation (different carving, mood, framing). - Not for: drawing a random pumpkin from a text description only, video, or editing footage. ## Steps 1. **Locate the reference image.** Use the image attached in this turn; otherwise the most recent relevant image in the conversation (an earlier upload, or a lantern image generated before). If there is no image at all, ask the user to upload one and stop — never invent a character and never generate from a written description alone. 2. **Read the carving motif off the reference.** Decide, silently, which 3–5 traits must survive into the pumpkin: head/hair silhouette, ears, horns, hat, crown or antenna shape; signature colour; accessory or prop; facial expression and mood. If the reference is ambiguous, keep the most graphic silhouette rather than inventing new features. 3. **Write the prompt in English.** Image prompts are always English, regardless of the language the user writes in. Default scene — use it verbatim when the user gives no extra direction: a roadside jack-o'-lantern suddenly lights up, and the character and the lantern look each other in the eye. Adapt only the bracketed parts: ```text Photorealistic cinematic night photograph, Halloween. Foreground: a real hand-carved jack-o'-lantern standing on a roadside, its carving designed after the reference character — [head / hair / horn / ear / hat silhouette] carved into the crown and ribs, [signature colour] lit from within, carved eyes and grin shaped like [character's expression], a real candle flame burning inside so the whole carving glows warm orange. [Character description] stands half a step away, turned toward the pumpkin and looking straight into its glowing eyes; the lantern has just flickered alight and its light spills across the character's face. Setting: quiet autumn roadside at dusk, wet asphalt and leaf litter, leaning wooden fence, bare tree silhouettes, faint fog, cobwebs, a few hanging string lights and scattered Halloween decorations. Lighting: warm orange key light from inside the pumpkin, cool blue-violet moonlight rim light, soft volumetric haze, warm glow spilling on the wet ground. Composition: eye-level, lantern in the lower third slightly off-centre, character in the middle third facing it, both in frame, 35mm lens look, shallow depth of field, the carving tack sharp, bokeh from the string lights. Style: film-photography realism, high detail, natural textures (pumpkin fibres, candle soot, fabric, skin), spooky-but-charming Halloween mood, no gore. No text, no watermark, no logo, not cartoon, not CGI, the pumpkin must be a real vegetable with a real flame, no melted or deformed carving, no extra limbs or fingers, no distorted face. ``` - Light must stay honest: the glow comes from the flame inside the pumpkin, not from an unexplained source. - Decorations, props and colour palette all stay on the Halloween theme the user asked for; never leave the scene an empty grey background. - If the user wants only the pumpkin, drop the character clause and make the carved lantern the sole subject (straight-on, chest-height, shallow depth of field). - If the user wants a change to an existing lantern image, pass that image as an extra reference and rewrite only the re