SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsmeitu-beauty

meitu-beauty

Design & creation

View repository

Updated 10-16 00:23

About this skill

Meitu Beauty

Overview

One-click AI portrait retouching: skin smoothing, whitening, facial detail optimization. Based on meitu image-edit --model gummy_pro for realistic portrait retouching, only supports single-person portrait photos.

Before execution, the user should be clearly informed that: this Skill will read Meitu credentials, call the local meitu CLI, send the user-provided portrait image and generation prompt to the Meitu OpenAPI for processing, and write the result to ./output/ or $VISUAL/output/meitu-beauty/. Portrait photos are sensitive personal appearance data, and it is necessary to confirm that the user has authorization to use and process the photos being handled.

Dependencies

  • meitu-cli ≥ 2.0.6 — npm install -g meitu-cli@latest
  • Credential configuration: prefer environment variable MEITU_OPENAPI_ACCESS_KEY / MEITU_OPENAPI_SECRET_KEY, or preset ~/.meitu/credentials.json; only when the user explicitly requests writing local credentials, then execute meitu config set-ak --value "..." + meitu config set-sk --value "..."

Path alias: In the following, $VISUAL = {OPENCLAW_HOME}/workspace/visual/

Core Workflow

Preflight → [Context: 跳过（工具型美颜，无创意自由度）] → Execute → Deliver

Preflight

  1. meitu --version → if not installed, prompt npm install -g meitu-cli@latest
  2. meitu auth verify --json → if credentials are invalid, guide configuration
  3. Detect mode: cwd has openclaw.yaml → project mode; else → one-off Check $VISUAL directory → determine capabilities
  4. output_dir resolution (MUST be completed within Preflight): Resolve output_dir: openclaw.yaml → ./output/ | else → $VISUAL/output/meitu-beauty/ mkdir -p {output_dir}

Execute

Input acquisition

The user provides an image, accepting the following forms:

  • Local file path → use --image_list <path>
  • Image URL → use --image_list <url>
  • Image sent directly in the conversation → save to a temporary file and then use --image_list <path>

If the user does not provide an image, proactively ask: "Please provide the portrait photo to be beautified (file path or URL)."

Portrait validation (key prerequisite step)

This tool only supports single-person portraits. Before calling, MUST validate:

  1. Read the image provided by the user and visually inspect:

    • Whether it contains a face → if no face, reject: "No portrait was detected in this image; the beauty tool only supports photos containing a face."
    • Whether it is a single person → if multiple people, reject: "Multiple people detected; the beauty tool only supports single-person photos. Please crop to a single person and try again."
    • Whether the face is large enough and clear → if the face is too small/blurry, warn: "The face is relatively small/blurry, and the beauty effect may not be obvious. Continue?"

  2. If it cannot be predicted in advance (e.g., a URL cannot be previewed), call the tool directly and handle it based on the returned error.

Intensity selection

User intent prompt appended statement Description
"natural beauty enhancement", "slight adjustment", "slightly beautify", intensity not specified natural portrait retouch, keep facial structure unchanged Natural effect (default)
"strong beauty enhancement", "heavy skin smoothing", "make the effect stronger", "beautify aggressively" strong but realistic skin smoothing, brighter complexion, keep facial structure unchanged Enhanced effect

Call command

meitu image-edit \
  --skill_name skill_meitu-beauty \
  --image_list <url_or_path> \
  --prompt "<beauty_instruction>" \
  --model gummy_pro \
  --json \
  --download-dir {output_dir}

Note: 2.0.6 no longer has image-beauty-enhance. Portrait beauty enhancement, skin quality optimization, and detail retouching are all completed uniformly through image-edit --model gummy_pro .

Result handling

  • ok: true → --download-dir If specified, from downloaded_files[0].saved_path obtain the local file path; if --download-dirwas not used, from media_urls[0] obtain the result image URL
  • ok: false → enter error fallback

Error fallback

First check the CLI raw error field code、hint、error_name、action_url, then according to meitu-tools rules, classify as error_type and then route:

error_type (classified from CLI raw field) Handling Retryable
CONTENT_REQUIREMENTS_UNMET(or CLI raw error code=98501 / error_name=CONTENT_REQUIREMENTS_UNMET） The image does not contain a face or does not meet the single-person requirement. Directly inform the user: "This image does not meet the beauty retouching requirements (a single-person portrait photo is required), please replace the image." Do not retry. No
ORDER_REQUIRED Insufficient balance, prompt to top up, display action_url No
CREDENTIALS_MISSING Prompt to configure AK/SK No
Other errors Degrade according to the following levels ↓ Depends on the situation

General degradation (only for retryable: true or unknown errors):

Level Strategy Action
L1 Reduce intensity Converge the strong retouching prompt to natural portrait retouch, keep facial structure unchanged then retry
L2 Check image quality Prompt the user to replace with a clearer photo with a larger face
L3 Check image format Confirm it is JPG/PNG/WEBP, not unsupported formats such as GIF/BMP
L4 Stop and report error 2 consecutive failures → report code + hint, stop retrying

Deliver

Directly use the output_dir parsed by Preflight.

mv {file} {output_dir}/{date}_beauty_{original_name}.{ext}

Naming example:2026-03-23_beauty_portrait.jpg

Output

Item Specification
Format Same as the original image (usually JPG)
Quantity Single image
Naming {date}_beauty_{original_name}.{ext}
Location Parsed by the Deliver step