According to AlphaMoat, as of 2026-09, Stable Diffusion has 2,687 downloads and 2 stars — ranked #3,957 of 5,078 AI skills overall, and #386 of 512 in Design & Media.

Installs

155

Downloads

2.7K

Stars

Overall Rank

#3,957

Top 78% overall · Top 76% in Design & Media

Platform	Installs	Downloads	Stars	Comments
ClawHub ↗	68	2.3K	1	-
SkillHub ↗	155	2.7K	2	—

Platforms count differently and update at different times.

What this skill does

It covers SD WebUI, ComfyUI, prompt engineering, LoRA training, sampler and parameter choices for planning image workflows. Copy-paste installation commands are included; review sources and versions before running them, prefer an isolated environment, and avoid elevated privileges. Actual generation needs a separately installed compatible model, frontend and CPU/GPU backend; training additionally needs suitable authorized images and training tooling. Enter cloned project directories explicitly before running installation commands. Model-specific LoRA, ControlNet, prompt syntax and parameters are not interchangeable; denoising controls here apply to image-to-image use. Treat hardware and sampler tables as planning guidance rather than guaranteed output or speed.

Profile published 2026-10-27

What you can do with it

Design prompts for image generation
Compare WebUI, ComfyUI or LoRA workflows

When to use it

When Stable Diffusion guidance is needed and installation sources have been reviewed

Example prompts

Generated from the skill’s documentation as examples; not results of actual runs.

Draft positive and negative prompts for an image theme without installing software.

List safety checks before running Stable Diffusion installation commands.

Monthly Trend

Downloads2.7K

Stars2

The counting method changed in 2026-09: skills with the same name are now counted separately by author, and the full catalog is covered. Dashed lines show months before the change; they are not compared with later months, and no growth is computed across the change.

Month	Downloads	MoM	Stars
2026-09	2.7K	—	2
2026-08	2.3K	+11.7%	2
2026-07	2.1K	+18.4%	1
2026-06	1.7K	+18.0%	2
2026-05	1.5K	—	1
2026-03	506	—	-

SKILL.md (original)

The author’s original file, shown as plain text without edits or review by AlphaMoat. Links are not clickable; check commands before running them. Fetched 2026-10-23.

3.6 KB · ClawHub v1.0.0

---
name: "Stable Diffusion"
version: "1.0.0"
description: "Stable Diffusion AI 绘画助手，精通 SD WebUI、ComfyUI、提示词工程、LoRA 训练"
tags: ["ai", "image", "stable-diffusion", "comfyui"]
author: "ClawSkills Team"
category: "ai"
---

# Stable Diffusion AI 绘画助手

Show all (99 lines)


你是 Stable Diffusion 领域的专家，精通图像生成的各个环节。

## 模型版本

| 模型 | 分辨率 | 特点 |
|------|--------|------|
| SD 1.5 | 512x512 | 生态最丰富，LoRA/插件最多，通用创作首选 |
| SDXL 1.0 | 1024x1024 | 画质大幅提升，双 CLIP 编码器，商业出图推荐 |
| SD 3 Medium | 1024x1024 | MMDiT 架构，文字渲染能力强 |
| SDXL Turbo | 512x512 | 蒸馏模型，1-4 步出图，实时预览 |
| Flux.1 | 最高 2048x2048 | Black Forest Labs 出品，指令遵循极强 |

## 前端工具

### SD WebUI (Automatic1111)
- 安装：`git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui && ./webui.sh`
- 优势：界面直观，插件生态成熟，扩展推荐 ControlNet、ADetailer、Tiled Diffusion

### ComfyUI
- 安装：`git clone https://github.com/comfyanonymous/ComfyUI && pip install -r requirements.txt`
- 优势：节点式工作流，可视化管线，适合复杂流程，可导出/导入 JSON 工作流

## 提示词工程

### 正向提示词结构
```
主体描述, 画质修饰, 风格标签, 光影氛围, 镜头语言
```
示例：`1girl, white dress, masterpiece, best quality, photorealistic, soft lighting, depth of field`

### 负向提示词（通用模板）
```
lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit,
fewer digits, cropped, worst quality, low quality, jpeg artifacts,
signature, watermark, blurry, deformed, ugly, duplicate
```

### 权重语法
- `(keyword:1.3)` — 增加权重到 1.3 倍
- `(keyword:0.7)` — 降低权重到 0.7 倍
- 嵌套：`((keyword))` 等价于 `(keyword:1.21)`

## 采样器选择指南

| 采样器 | 步数建议 | 特点 |
|--------|----------|------|
| Euler a | 20-30 | 速度快，创意性强，结果多样 |
| DPM++ 2M Karras | 20-30 | 画质稳定，细节丰富，推荐通用 |
| DPM++ SDE Karras | 20-30 | 细节最佳，适合写实风格 |
| DDIM | 20-50 | 确定性强，适合 img2img |
| UniPC | 15-25 | 收敛快，少步数即可出好图 |
| LCM | 4-8 | 极速采样，需配合 LCM LoRA |

## 关键参数

| 参数 | 推荐范围 | 说明 |
|------|----------|------|
| CFG Scale | 5-12 | 提示词引导强度，7 为通用值，过高会过饱和 |
| Steps | 20-40 | 采样步数，越多越精细但速度越慢 |
| Seed | -1 或固定值 | -1 随机，固定值可复现结果 |
| Denoising | 0.3-0.7 | 仅 img2img，越高变化越大 |
| Clip Skip | 1-2 | SD1.5 动漫风建议 2，写实建议 1 |

## LoRA / ControlNet

### LoRA 训练要点
- 数据集：20-50 张高质量图片，统一风格和分辨率
- 工具：kohya_ss GUI 或 `accelerate launch train_network.py`
- 关键参数：`network_rank=32`、`learning_rate=1e-4`、`epochs=10-20`
- 使用：提示词中加 `<lora:模型名:权重>` 触发，权重建议 0.6-0.9

### ControlNet 控制类型
- Canny：边缘检测，精确控制轮廓
- OpenPose：人体姿态控制
- Depth：深度图控制空间关系
- Tile：超分辨率和细节增强
- IP-Adapter：图像风格迁移，以图生图的高级方案

## 硬件需求

| 配置 | 显存 | 适用模型 |
|------|------|----------|
| 入门 | 6GB (GTX 1660) | SD 1.5 基础出图 |
| 推荐 | 8-12GB (RTX 3060/3080) | SD 1.5 全功能 + SDXL |
| 高端 | 16-24GB (RTX 4080/4090) | SDXL + ControlNet + 大批量 |

Mac M 系列通过 MPS 后端支持，M2 Pro 以上体验尚可。云端推荐 AutoDL（国内）或 RunPod。