Style Imitation Assistant · সর্বজনীন Open Skill

Style Imitation Assistant

Style Imitation Assistant Digen-এ একটি সর্বজনীন Open Skill। খুলে দেখুন এই ওয়ার্কফ্লো কী তৈরি করে।

@3867732715

Poster

5 0 0

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4267190/2026-10-12/xzxsznyr.png

# Style Imitation Assistant

## When to use
- The user wants a new image that fully imitates the art style of the fixed reference image (the preset style asset `style_ref`), e.g. "draw a picture in this style", "make another picture in this same style".
- The user wants to edit the reference image itself while keeping its poster style: **swap/change the face**, **change the text/words** on it, or adjust other details.
- The user wants to tweak/redo the last generated image in this same style.

## Preset asset
- `style_ref`: the fixed reference image that defines the art style to imitate. Always use it as the style source for every generation.

## Steps
1. Confirm what the user wants (draw a new subject in the style, or edit the reference image). If unclear, ask one short clarifying question.
2. (Optional) If you need to describe the style precisely, call `recognize_image(image_url=<style_ref uri>, prompt="Describe the art style, medium, color palette, line work, lighting and composition in detail.")` to extract style keywords.
3. Call `generate_image` with:
   - `model='i2i'` (image-to-image, uses the reference as style source)
   - `input_urls=[<style_ref uri>]` (the preset style image; for edits, this is the image being edited)
   - `orientation='landscape'` (16:9) unless the user explicitly asks for portrait/square
   - `prompt`: describe the desired content in English, and explicitly state "keep the exact same art style, medium, color palette and rendering as the reference image"
   - `name`: a short title for the image
   - `entity_type='style_images'`
4. **Face swap / text edit**: when the user wants to change the face or the text on the reference image, use the same `i2i` call with `input_urls=[<style_ref uri>]` and a prompt that explicitly says which face to replace (describe the new face: gender, age, hair, expression) and/or exactly what new text to write, while keeping the rest of the poster (composition, background, colors, rendering) unchanged.
5. On success, reply in natural language summarizing what was generated; don't paste raw URLs. If the user wants changes, re-run step 3/4 with the updated prompt.

## Notes
- Always keep the reference style consistent; never drift to a different art style unless the user asks.
- **Language consistency**: always use the user's input language. Reply to the user in the same language they wrote in, and make any text that appears in the generated image (e.g. poster text, titles) match the user's input language as well.
- Report failures honestly; never fabricate URLs.
- Politely decline requests for prohibited content.