OC generation · публичный навык Open Skill
OC generation
OC generation — публичный Open Skill на Digen. Откройте, чтобы увидеть, что создаёт этот процесс.
Manhua
2 2 5
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3947739/2026-08-25/ef2fm984.jpg
# OC Character Card Generator ## When to use - The user wants to create an "OC character card": a full-body portrait of the character on the left with a text introduction on the right, on a pure white background. - The user has uploaded a character image (wishing to keep that character's design and art style), or only provided a text description (wishing to generate a character from the description). - The user also wants a three-view sheet (front/side/back) of the character. ## Language rule - Communicate with the user in the language the user inputs, and keep all generated content in the same language as the user's input: the character profile (step 1) and the text on the OC card (step 5) must be written in the user's input language. If the user inputs Chinese, generate Chinese text; if English, generate English text; and so on for any other language. ## Steps ### 1. Collect character information - Ask the user for the character's basic information: name, age, identity/occupation, personality, backstory, etc.; if the user only gave scattered details or none at all, automatically complete a reasonable character profile (name, age, identity, 2-3 personality keywords, a one-line bio, backstory), written in the user's input language. - If the user uploaded an image, first use `recognize_image` to analyze it and extract: appearance details (hairstyle, hair color, eye color, face shape, outfit, accessories, vibe) and art style traits (e.g., anime, semi-realistic CG, Chinese-style thick paint, chibi, watercolor, etc.), as the basis for the subsequent redraw and copy. - Show the generated/completed character profile (including the backstory) to the user and ask if they are satisfied; if not, revise the profile according to their feedback until the user confirms, then proceed to the next step. ### 2. Determine the art style - If the user uploaded an image: the art style follows the original image, and the redraw must keep the art style exactly the same as the original. - If the user has no image: first ask which art style they want (e.g., anime, semi-realistic CG, Chinese-style thick paint, chibi, watercolor, etc.), and continue once answered. ### 3. Generate the full-body portrait - If the user uploaded an image: call `generate_image`, `model='i2i.hd.lite'`, `input_urls=[user's uploaded image]`. Prompt requirements: keep the original character's design and art style completely unchanged, redraw as a full-body standing portrait (full body in frame, front-facing or slightly turned, natural pose, slender waist and long legs, elongated body proportions), with a soft solid-color gradient background (e.g., light-toned gradient/watercolor wash effect), clean and neat. - If the user has no image: call `generate_image`, `model='t2i'`. Generate a full-body standing portrait from the user's text description + the chosen art style (full body in frame, front-facing or slightly turned, natural pose, slender waist and long legs, elongated body proportions), with a soft solid-color gradient background (e.g., light-toned gradient/watercolor wash effect). - Parameters: `orientation='portrait'`, `resolution='720P'`, `name='OC Full-body Portrait'`, `entity_type='oc_portraits'`. - Show the generated full-body portrait to the user and ask if they are satisfied; if not (design, art style, pose, body proportions, gradient color, etc.), regenerate or adjust according to their feedback until the user confirms, then proceed to the next step. ### 4. Generate the three-view sheet - Using the full-body portrait confirmed in step 3 as the reference image, call `generate_image`, `model='i2i.hd.lite'`, `input_urls=[full-body portrait]`. - Prompt requirements: same character, same art style. Composition: on the left is a close-up/headshot of the character above the chest (Close-up / Headshot), taking up about 45-50%; on the right is a turnaround view arrangement (Turnaround View), taking up about 50-55%, from left to right: front view (Front View), 90° side v