Global Airline Uniform Portrait — Open Skill
Global Airline Uniform Portrait
Global Airline Uniform Portrait is a public Open Skill on Digen. Open it to see what this workflow creates.
Product
0 0 0
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4284865/2026-10-22/y6mwq5yk.jpg
# Global Airline Uniform Portrait Turns a person into a realistic flight-attendant portrait. The uploaded identity must survive unchanged, while the uniform, posture and cabin scene follow the requested airline theme. Output is still images only (no video, no audio). ## When to use - The user uploads one or more portrait photos and asks for a "flight attendant", "cabin crew" or "airline uniform" version of that person. - The user names an airline (or a uniform color / country) and wants the corresponding uniform portrait. - The user wants to redo only one variable of an earlier result: another airline, another uniform color, another pose or framing. - The user gives no photo and only describes the person or the airline in text. ## Inputs and outputs - Input: text brief (airline, color, framing, mood, number of variants) plus optional reference photo(s). - Output: one or more images produced with `generate_image`. - Always pass a short `name` (e.g. `Airline uniform portrait - <airline>`) and `entity_type='characters'`. - Channel choice: - Reference photo available (the normal case, includes uniform swap): `generate_image(..., input_urls=[reference photo], model='i2i.hd.lite')`. - No reference photo, text only: `generate_image(..., model='t2i')`. - Default framing for this app: `orientation='portrait'`, `resolution` left default (720P) unless the user asks for higher quality. ## Steps 1. **Brief check.** Confirm four things from the conversation: the person (photo or description), the airline or uniform theme, the framing (full body / three-quarter / close), and how many variants. If a reference photo exists, treat it as the identity source and say so in one short sentence. Do not ask for anything already implied by the photo or the previous turn. 2. **Read the stage plan once.** Call `read_skill_file(name='airline-uniform-portrait', path='references/workflow-stages.md')` to get the stage table, the channel for each stage, and the exit criteria. If the task spans more than one stage, call `write_todos` with those stages before continuing. 3. **Profile the uniform.** If the user named an airline whose uniform you are not sure about, collect the real palette and signature pieces with `search_web` (query like `<airline> cabin crew uniform color scarf`). If nothing reliable is found, keep the description at palette + generic cabin-crew level and tell the user the uniform is an approximation. Never invent official details, tails, or logos. 4. **Optional identity read.** When the reference photo is low resolution, ambiguous, or contains several people, call `recognize_image` on it to record which person is meant and which identity cues must be preserved. Skip this if the photo is already clear. 5. **Build the prompt from the reference file.** Immediately before generating, call `read_skill_file(name='airline-uniform-portrait', path='references/style-and-prompt.md')` and follow it: fill the identity block, the uniform block, the scene block, the pose block, the camera/light block, the quality block and the negative block with the current user intent. Do not generate with a generic one-line prompt. 6. **Generate.** - With a reference photo: `generate_image(prompt=<assembled prompt>, input_urls=[reference photo], model='i2i.hd.lite', orientation='portrait', name='Airline uniform portrait - <airline>', entity_type='characters')`. - Text only: `generate_image(prompt=<assembled prompt>, model='t2i', orientation='portrait', name='Airline uniform portrait - <airline>', entity_type='characters')`. - One image per call; for several variants change exactly one variable between calls (framing, pose, or scene) and keep the identity block identical. 7. **Review and retry.** Apply the review checklist in `references/step-details.md` to the returned image (identity match, uniform theme, anatomy, cabin plausibility, tastefulness). If a check fails, fix only the failing block of the prompt and regenerate, at most twice; if it