Updated: Jul 9, 2026
action
Simply enter the batch of images you want to generate and press "Run" to produce a variety of images featuring diverse poses.
Here is a breakdown of its Randomization Features 🎲
Hey everyone! 👋 I just shared my latest ComfyUI workflow. If you enjoy highly dynamic and unpredictable generations, this workflow is built entirely around randomization. Here is how the "gacha" mechanics work under the hood:
📐 Dynamic Aspect Ratios via Probability Gates
The workflow doesn't use a fixed resolution! It features a logic system using ProbabilityGate nodes (set to a 33% chance with randomized seeds) linked to latent switches.
Every generation will randomly "roll" and choose between different canvas sizes and orientations (e.g., Portrait, Landscape, or Square) on the fly.
🃏 Massive "Gacha-Style" Dynamic Prompt System
This is the core of the workflow! It uses a massive network of text nodes to build a completely unique prompt for every single generation:
Wildcards: Uses WildPromptor and easy wildcards (with randomized seeds) to randomly draw base character traits like age, hairstyle, hair color, eye color, and outfits.
Weighted Text Routing: Uses batch_randoml_text nodes to randomly select different camera angles and core actions based on predefined weights.
Conditional Probability Traits: It contains dozens of BatchConditionalTextWeight nodes grouped by categories (Expressions, Poses, Actions, etc.). Each specific tag (like blushing, sweatdrops, or specific interactions) has its own independent trigger probability (e.g., 5%, 10%, 30%) and a randomized seed.
The Result: Every time you generate an image, the workflow basically rolls the dice for every single detail of the character, action, and scene. It's an automated prompt gacha machine!
Feel free to download it, try it out
If you have any questions about usage, you can contact me on Discord.
sample images👇
https://www.pixiv.net/artworks/146632813