Scrape WeChat Group Avatars and Rank Them with Python & Baidu AI

Inspired by a class group chat, the author decided to rank classmates' profile pictures by perceived attractiveness using Python.

1. Download group avatars

The author chose to avoid WeChat SDKs like itchat or wxpy and instead accessed the web version of WeChat. After scanning the QR code and logging in, the server returns user information in three POST requests, each delivering up to 50 user records (nickname, group nickname, avatar URL, gender, region, etc.).

By reproducing the request payload found in the web traffic, the script can fetch the complete list of members and extract each avatar URL.

2. Save avatars locally

Using the same session, the script iterates over the avatar URLs and downloads each image, naming the files with the corresponding group nickname. The downloaded images are stored in a local img folder.

3. Register Baidu AI and obtain credentials

Sign up for a Baidu Cloud account, create a Face Recognition application, and note the API Key and Secret Key , which are required to request an access_token.

4. Build the token request function

The script defines a get_access_token function that sends a request with the API Key and Secret Key and receives the access_token needed for subsequent face‑analysis calls.

5. Analyze each avatar

For every saved image, the script calls Baidu's face detection endpoint, receiving attributes such as age, gender, facial shape, smile intensity, and a beauty score. The author keeps only gender, age, and beauty for ranking.

6. Filter and rank results

After obtaining the analysis, the script discards non‑female faces, saves the remaining data (including beauty scores) into a spreadsheet, and produces a ranking of the female avatars based on the AI‑generated scores.

The final ranking was shared back to the WeChat group, illustrating a playful yet technically detailed use of web scraping and AI face analysis.