Gothic Palace · Candlelit Film Still · публичный навык Open Skill
Gothic Palace · Candlelit Film Still
Create a cinematic moment from your own character photo. Upload one clear adult reference, choose a pose and frame shape, and step into a Gothic palace filled with ivory lace, antique mirrors and golden candlelight. The scene captures a natural turn toward someone approaching, with layered foregrounds and realistic skin texture. How to use: Upload your character photo, optionally request 3:4 or 16:9, then ask for a candlelit palace film still. Your photo controls the identity; the example supplies the styling. Example provenance: The cover is a RunningHub template test, originally 2480 × 3312 pixels. DIGEN generation testing has not yet passed; results and resolution depend on the actual task returned by the configured image channel.
Film
1 0 2
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4274163/2026-10-14/ewq39h3p.jpg
# Gothic Palace - Candlelit Film Still ## When to use Turn one adult character photograph into a single cinematic moment inside a candlelit Gothic palace: ivory lace, an embroidered veil, antique gold mirrors, dark velvet and warm candlelight. Preserve the supplied person's identity and create a natural narrative moment caught mid-turn. ## Required input One clear photograph of an adult character. Ask for it if missing or unreadable. Do not use the cover as a substitute for the user's identity reference. Optional choices: pose and aspect ratio (3:4 by default, or 16:9). ## Reference file Read references/prompt-template.md before generation. Keep the scene and lighting while adapting identity only from the user's photograph. ## Style example https://rh-canvas-files.xiaoyaoyou.com/acaf1d02a918f6131defede49c8da998/output/00cf6f61-9469-44c5-b966-94639d2e6dc5.png Use this image for clothing, setting and light only. It was produced in RunningHub; it is not evidence of a completed DIGEN generation. ## Workflow 1. Confirm the identity photograph is available and readable. Preserve visible facial structure, proportions, hair and eye color, and natural skin details. Do not infer nationality from appearance or impose the example character's face. 2. Build the English generation prompt from references/prompt-template.md. Keep the natural turning action and layered cinematic composition. 3. Call generate_image with model='i2i.hd.gpt', input_urls=[identity_photo, style_example], resolution='4K' when supported, name='Gothic Palace - Candlelit Film Still', and entity_type='characters'. Set the requested aspect ratio using the tool's current supported parameters: 3:4 by default, 16:9 when requested. If exact ratio is unsupported, report the limitation instead of claiming that landscape orientation guarantees it. 4. Inspect the output with recognize_image for identity consistency, natural hands, lace texture, candlelight, narrative composition and actual pixel dimensions. Retry at most once for a clear visual failure, using the same model and identity reference. Report unresolved problems. 5. Display the generated image and the real download URL returned by the platform; use get_presigned_url if required and available. Report actual dimensions and any material limitation. Never invent a result or URL. ## Model and provenance The configured DIGEN channel is i2i.hd.gpt, reported by the platform as gpt-image25. This label does not prove equivalence to another provider's model. The source RunningHub canvas reports G2 official stable; its exact underlying version and cross-platform equivalence remain unverified. Do not claim a matched-model validation or successful DIGEN test without a completed task record. If the configured channel is unavailable, stop and explain; never silently switch to i2i.hd, Nano Banana, Fast, Mini or another version. ## Output rules Generate one still image, not video. Keep the user's chosen identity; the example face is never a default override. Do not add Midjourney-specific flags to this tool. A requested 4K setting is not proof of native 4K output: report returned dimensions. All listing copy, instructions and prompt text are English.