chengyansen-ai/cys-migration-skill — explained in plain English
Analysis updated 2026-05-18
Generate large batches of unique, realistic AI portrait photos for a social media account.
Fix common AI body proportion mistakes like oversized heads in full body shots.
Apply one of 30 plus style presets, from historical Chinese fashion to modern cyberpunk looks.
Check generated prompts against platform content rules before posting images.
| chengyansen-ai/cys-migration-skill | 0pen-sourcer/hearth | 13127905/deep-learning-based-air-gesture-text-recognition- | |
|---|---|---|---|
| Stars | 15 | 15 | 15 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 2/5 | 3/5 | 3/5 |
| Audience | vibe coder | vibe coder | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires an AI image generation platform such as ComfyUI or Krea2 to actually produce images, this repo only supplies prompts and scripts.
This project is a prompt writing method for creating realistic looking AI generated portrait photos, built from testing over 5,000 real generated images. It is aimed at people who use AI image tools such as Coze, Dify, Claude, Codex, ComfyUI, Krea2, or GPT Image 2, and it packages up a repeatable recipe rather than a piece of software you run. The core idea is a nine part Chinese language prompt template that separates a person's identity, face, makeup, body, accessories, pose, clothing, environment, and camera framing into distinct sections that can be reused across many generated images. It also includes strict rules for things like keeping the head no more than a quarter of the frame and the legs at least two thirds of it, so that full body shots come out with correct human proportions instead of the common AI mistake of an oversized head. Alongside the template are recommended settings for the Krea2 image generation pipeline, such as fixed creativity settings and limits on how strongly a character style file should be applied. There are more than 30 ready made style presets covering historical Chinese fashion eras as well as modern and international looks such as cyberpunk, Korean fashion, or gothic. The repository also ships two small Python scripts that can batch generate large sets of unique prompt variations, for example one thousand full body images or half body images at once, along with a self test script that checks the generated prompts for internal contradictions and against a list of platform content rules for a Chinese social platform. In short, this is a structured, tested prompt engineering toolkit for people producing large volumes of realistic AI portrait images, focused on correct body proportions, consistent style, and content compliance, rather than a traditional coding library. It is released under the MIT license.
A tested prompt template and small batch generation scripts for producing realistic AI portrait photos with correct body proportions and style variety.
Mainly Python. The stack also includes Python, ComfyUI, Krea2.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.