---
title: "Product Review | Chushou AI Pro: Shared Open Creation + Model Upload, Train & Invoke"
author: "Unique Research"
sourcePublication: "Unique Research (FFCAP English library)"
originalPublishedAt: "2023-07-04T16:42:20+08:00"
canonical: "https://ffcap.cn/en/research/20230704a05y1100"
source: "https://ffcap.cn/en/research/20230704a05y1100"
language: "en"
cnSourceUrl: "https://view.inews.qq.com/a/20230704A05Y1100"
---

# Product Review | Chushou AI Pro: Shared Open Creation + Model Upload, Train & Invoke

_Unique Research · Insight / product review · July 4, 2023_

**Chushou AI Pro** (触手AI专业版) is the AI painting product of Jellyfish Intelligence (水母智能), itself incubated by LKK Innovation Design Group (洛可可). Available as an app and as mini-programs on WeChat / Douyin / Kuaishou, it aims to build a shared open creation ecosystem for model trainers, creators, and hobbyists — claiming **millions of users** and **100M+** AIGC works generated. This review covers the web Pro edition then in public beta at **https://acgnai.com/** (Edge 114, June 2023).

## Creation

Users generate images from text, images, or reference images.

## Text-to-image

Enter a short keyword prompt → choose style, aspect ratio, and count → generate. **UX highs:** (1) broad style models — system-recommended, community-shared, and user-trained — with **model stacking** for niche looks; (2) post-gen actions (params, regen, variation, HD, img2img, reference, download, add to works); (3) prompt engineering with recommended positive/negative terms to lower the barrier. **Friction:** many styles without preview tiles — suggest sample outputs per style.

## Image-to-image

Upload → “magic brush” local mask → keyword for the region → style → generate. Side-by-side original vs output and the same quick actions help iteration. Mask quality can miss expectation; algorithm/UI polish needed.

## Reference generation

Reference image + text → new image. Condition options: pose, region, depth, contour strength, normal maps, line maps — strong ControlNet-style flexibility. Parameter panel sits mid-lower in the workspace (hard to find); shortcuts or a split model/params workspace would help.

## Training — private models

No local GPU required: upload ≥8 images, set fast-mode params, train a private model. Training data can be local uploads or prior generated works — a clear differentiation for personalized styles.

## SWOT & commercial read

-   **Demand:** digital art / design demand + pro tools (private training, model stack) should keep pull among pros and enthusiasts.
-   **Moat:** feature completeness and pro focus; community models + private train as sticky loops.
-   **Monetization:** premium features, private training, community/ads/partnerships.
-   **Risks:** crowded AI drawing market, tech shifts, shifting user expectations — continuous iteration required.

_8 CN images (structure / feature screens); attach on ingest. Author Xiao Heng · Review Ski, Huajia._

— END —

---

Editorial English edition published on UniqueCapital: https://ffcap.cn/en/research/20230704a05y1100
On-site reading page: https://ffcap.cn/en/research/20230704a05y1100