Skip to content

acer1204/CivitAI-Manager-Plus

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CivitAI Manager Plus

A reliable, streamlined CivitAI model manager extension for Stable Diffusion WebUI. As a frequent CivitAI user, I referenced existing tools and built one that's smooth to use and actually stable.

中文說明

Why This Extension?

After testing many existing CivitAI extensions, I found common issues that never seem to get fixed:

  1. Overcomplicated UI - Too many features crammed together, making simple tasks difficult
  2. Poor offline support - Saved model info breaks or shows errors when models are taken down from CivitAI
  3. Redundant network requests - Even for already-downloaded models, browsing still fetches data from the internet every time, making it painfully slow
  4. Unstable batch downloads - Batch downloading frequently freezes or gets stuck on loading
  5. Manual path management - Having to manually set download paths for every model

What CivitAI Manager Plus does differently:

  1. Simplified UI - Clean interface, automation over manual steps. Click a card to view info, check boxes to batch download
  2. True offline model info - Once saved locally, model info (description, all preview images, prompts) works completely offline. Even if a model is removed from CivitAI, your saved info remains intact and looks identical
  3. Local-first browsing - Installed models load previews from local .preview.png/.mp4 files instantly. Only fetches from CivitAI when no local preview exists (and caches it for next time)
  4. Async batch downloads - Non-blocking background downloads. Queue models from multiple pages without freezing the UI
  5. Auto-organized paths - Downloads automatically sorted into BaseModel / Author / ModelName folders

Screenshots

Browse & Search

Browse Tab

Model Info - Description

Model Info Description

Model Info - Preview Images with Prompts

Model Info Preview

Installed Models with Folder Tree

Installed Tab

Local Cached Model Info

Installed Model Info

Local Cached Preview Images

Installed Preview

Download Queue

Downloads Tab


Features Overview

Browse Tab - Search & Discovery

  • Flexible Search - Search by model name, username, or tag
  • Advanced Filters - Filter by content type (Checkpoint, LORA, LoCon, DoRA, VAE, ControlNet, etc.), base model (SD 1.5, SDXL, Pony, Illustrious, NoobAI, Flux.1, SD 3.5, etc.), sort order, and time period
  • NSFW Toggle - Show or hide NSFW content
  • Adjustable Page Size - 5 to 100 cards per page
  • Cursor-Based Pagination - Reliable page navigation for large result sets
  • Installed Indicator - Green badge on models you already have
  • Search Config Persistence - All search settings (including Search by, Show NSFW, Save Local Info) are saved and restored across sessions

Model Cards

  • Thumbnail preview with lazy loading
  • Video thumbnail support (auto-play)
  • Model name, base model, and metadata
  • Selection checkbox for batch operations
  • Click card to open Model Info popup
  • Hover animation with accent glow

Batch Download

  • Select All / Deselect All - Quickly select models (auto-skips already installed)
  • Download Selected - Queue all selected models for background download
  • Save Local Info - Optionally cache model info and preview images locally during batch download
  • Non-blocking - Downloads run in background, you can continue browsing and adding more

Model Info Popup

Full-screen popup with detailed model information, accessible from both Browse and Installed tabs by clicking any model card.

Header:

  • Model title linked to CivitAI page
  • Creator name and content type
  • Data source badge: Online (blue) / Local (green)
  • Save Local checkbox to cache all data offline

Description Tab:

  • Full model description with embedded images
  • Scrollable content within the popup

Preview Images Tab:

  • Grouped by version (up to 5 versions)
  • Each image shows:
    • Preview thumbnail (image or video)
    • Prompt and negative prompt with Copy buttons
    • Generation parameters (steps, sampler, CFG scale, seed, size, model, clip skip)
    • Send to txt2img button - sends all generation parameters directly to txt2img

Save Local (Offline Model Info)

Save complete model information locally for offline access. Useful when models may be taken down from CivitAI.

What gets saved:

  • Model metadata (name, creator, type, description)
  • All preview images and videos (downloaded to local disk)
  • Images embedded in model description
  • Generation parameters (prompts, settings) for each preview image

How it works:

  • Check the Save Local checkbox in the model info popup, or
  • Enable the Save Local Info checkbox in Browse tab before batch downloading
  • Saved data is stored in model_info_cache/{model_id}/
  • When reopening model info, local data is used automatically (no API calls)
  • Badge shows Local (green) when viewing cached data

Favorites Tab

  • Save frequently used models for quick access
  • Card grid display with thumbnails
  • Click to view model details
  • Persistent storage in civitai_favorites.json

Installed Models Tab

Sort & Folder Tree (Left Panel):

  • Sort by dropdown: Publish Date (Newest/Oldest), Filename (A-Z/Z-A) - persisted across sessions
  • Hierarchical folder navigation based on actual disk structure
  • Shows models/ and embeddings/ directories
  • Click any folder to directly load and display its models

Model Grid (Right Panel):

  • Shows installed models with thumbnails (image and video)
  • Local preview files (.preview.png, .preview.mp4) loaded instantly
  • Missing previews fetched in background and cached for next time
  • File size and folder path info
  • Click to open model info popup (if CivitAI metadata available)

Saved Metadata per Model:

  • Publish date, trigger words, download date, model name, version name, SHA256

Downloads Tab

  • Download Queue with live progress display (auto-refresh)
  • Progress Bar with download speed and ETA
  • Concurrent Downloads - Download multiple models simultaneously (configurable 1-5)
  • Auto-Retry - Up to 5 attempts with adaptive backoff for CDN errors
  • Retry Failed button - Re-queue all failed downloads with one click
  • Metadata Auto-Save - .json with publish date, trigger words, SHA256 saved alongside each model
  • Preview Auto-Save - .preview.png or .preview.mp4 saved alongside each model
  • Background Processing - Downloads continue even while browsing other tabs
  • Clear Finished - Remove completed/failed items from the queue

Settings

Configurable in Settings > CivitAI Manager Plus:

Setting Default Description
CivitAI API Key (empty) Personal API key for authenticated downloads
Auto-organize Downloads ON Organize by base_model / author / model_name
Max Concurrent Downloads 2 Parallel download limit (1-5)
Use Aria2 ON Faster, resumable downloads (fallback to HTTP)
Show NSFW by Default OFF Initial NSFW content visibility
HTTP Proxy (empty) Proxy URL (e.g., http://127.0.0.1:7890)

Installation

Method 1: From URL (Recommended)

  1. Open Extensions tab in SD WebUI
  2. Go to Install from URL
  3. Paste: https://github.com/acer1204/SD-CivitAI-Manager-Plus.git
  4. Click Install
  5. Restart SD WebUI

Method 2: Manual

cd stable-diffusion-webui/extensions
git clone https://github.com/acer1204/SD-CivitAI-Manager-Plus.git

Restart SD WebUI.


Usage

  1. Navigate to the CivitAI Manager Plus tab in the WebUI
  2. Enter a search query and click Search (or press Enter)
  3. Use the Filters accordion to narrow results
  4. Click a model card to view detailed model info popup
  5. For batch downloads: check model cards, then click Download Selected
  6. Check Save Local Info to cache model info offline
  7. Go to Downloads tab to monitor progress

Requirements

  • Stable Diffusion WebUI (AUTOMATIC1111 or compatible fork)
  • Python 3.10+
  • requests library (auto-installed)
  • Optional: aria2c for accelerated downloads
  • Optional: send2trash for safe file deletion

License

This project is licensed under the AGPL-3.0 License.


中文說明

CivitAI Manager Plus

穩定、精簡的 CivitAI 模型管理擴充功能,適用於 Stable Diffusion WebUI。自己常用 CivitAI,參考了現有的工具後,順手寫了一個自己用起來順手也穩定的擴充。

為什麼開發這個擴充?

測試了許多現有的 CivitAI 擴充功能後,發現以下常見問題始終沒有被解決:

  1. 介面太雜不好操作 - 功能堆疊過多,簡單的操作反而變得複雜
  2. 離線保存方式太陽春 - 模型被 CivitAI 下架後,保存的資訊不是顯示異常就是直接壞掉
  3. 明明已離線下載,瀏覽時仍然連網抓取 - 即使模型已下載到本地,每次瀏覽還是重新從網路抓資料,導致瀏覽效率極低
  4. 批量下載容易當機 - 批量下載經常卡在讀取中或直接凍結
  5. 下載路徑要手動設定 - 每個模型都要自己改路徑,非常麻煩

CivitAI Manager Plus 的改良:

  1. 簡化操作,自動化取代手動 - 乾淨的介面,點擊卡片查看資訊,勾選即可批量下載
  2. 真正的離線模型資訊 - 勾選本地保存後,模型資訊(說明、所有預覽圖、提示詞)完全離線可用。即使模型從 CivitAI 下架,已保存的資訊完整不受影響,預覽視窗看不出差異
  3. 本地優先瀏覽 - 已安裝模型直接從本地 .preview.png/.mp4 檔案載入預覽圖,瞬間顯示。只有在沒有本地預覽時才從 CivitAI 抓取(並快取供下次使用)
  4. 批量下載改為非同步處理 - 背景下載不阻塞介面,可以跨頁繼續選取模型加入佇列
  5. 自動整理下載路徑 - 下載自動依 基底模型 / 作者 / 模型名稱 分類存放

截圖展示

瀏覽與搜尋

Browse 分頁

模型資訊 - 說明

模型資訊說明

模型資訊 - 預覽圖與提示詞

模型資訊預覽

已安裝模型與資料夾樹

Installed 分頁

本地快取的模型資訊

已安裝模型資訊

本地快取的預覽圖

已安裝預覽圖

下載佇列

Downloads 分頁


功能概覽

Browse 分頁 - 搜尋與發現

  • 彈性搜尋 - 支援以模型名稱、使用者名稱或標籤搜尋
  • 進階篩選 - 依內容類型(Checkpoint、LORA、LoCon、DoRA、VAE、ControlNet 等)、基底模型(SD 1.5、SDXL、Pony、Illustrious、NoobAI、Flux.1、SD 3.5 等)、排序方式及時間範圍篩選
  • NSFW 開關 - 顯示或隱藏 NSFW 內容
  • 可調每頁數量 - 每頁 5 至 100 張卡片
  • 游標分頁 - 大量結果也能可靠翻頁
  • 已安裝標示 - 已下載的模型顯示綠色徽章
  • 搜尋設定記憶 - 所有搜尋設定(含 Search by、Show NSFW、Save Local Info)重啟後自動還原

模型卡片

  • 縮圖預覽(支援圖片和影片自動播放)
  • 模型名稱、基底模型等資訊
  • 勾選框供批量操作
  • 點擊卡片直接開啟 Model Info 彈窗
  • 滑鼠懸停動畫效果

批量下載

  • 全選 / 取消全選 - 快速選取(自動跳過已安裝)
  • 下載已選 - 將所有選取的模型加入背景下載佇列
  • 本地保存資訊 - 可選擇在批量下載時同時快取模型資訊和預覽圖
  • 非阻塞 - 下載在背景執行,可繼續瀏覽和新增下載

Model Info 彈窗

全螢幕彈窗,顯示詳細模型資訊。在 Browse 和 Installed 分頁中點擊任何模型卡片即可開啟。

標題區:

  • 模型名稱(連結至 CivitAI 頁面)
  • 作者與內容類型
  • 資料來源徽章:Online(藍色)/ Local(綠色)
  • Save Local 勾選框,一鍵離線快取所有資料

Description 分頁:

  • 完整模型說明(含嵌入圖片)
  • 可捲動的內容區域

Preview Images 分頁:

  • 依版本分組(最多 5 個版本)
  • 每張圖片顯示:
    • 預覽縮圖(圖片或影片)
    • 正向提示詞與反向提示詞,附 Copy 按鈕
    • 生成參數(steps、sampler、CFG scale、seed、size、model、clip skip)
    • Send to txt2img 按鈕 - 將所有生成參數直接傳送至 txt2img

本地保存(離線模型資訊)

將完整模型資訊保存至本地,供離線存取。適用於模型可能從 CivitAI 下架的情況。

保存內容:

  • 模型中繼資料(名稱、作者、類型、說明)
  • 所有預覽圖片和影片(下載至本地磁碟)
  • 說明中嵌入的圖片
  • 每張預覽圖的生成參數(提示詞、設定)

使用方式:

  • 在 Model Info 彈窗中勾選 Save Local,或
  • 在 Browse 分頁批量下載前勾選 Save Local Info
  • 資料儲存於 model_info_cache/{model_id}/
  • 重新開啟模型資訊時自動使用本地資料(不需 API 呼叫)
  • 使用本地資料時徽章顯示 Local(綠色)

Favorites 分頁

  • 收藏常用模型以便快速存取
  • 卡片格線顯示,附縮圖
  • 點擊查看模型詳情
  • 持久儲存於 civitai_favorites.json

Installed 分頁 - 已安裝模型

排序與資料夾樹(左側面板):

  • 排序 下拉選單:發布日期(最新/最舊)、檔名(A-Z/Z-A)— 設定會保存
  • 階層式資料夾導覽,反映實際磁碟結構
  • 顯示 models/embeddings/ 目錄
  • 點擊任何資料夾直接載入並顯示其中的模型

模型格線(右側面板):

  • 顯示已安裝模型及縮圖(支援圖片和影片)
  • 本地預覽檔案(.preview.png.preview.mp4)瞬間載入
  • 缺少的預覽圖在背景自動抓取並快取
  • 顯示檔案大小和資料夾路徑
  • 點擊開啟 Model Info 彈窗(需有 CivitAI 中繼資料)

每個模型保存的中繼資料:

  • 發布日期、觸發詞、下載日期、模型名稱、版本名稱、SHA256

Downloads 分頁

  • 下載佇列 即時進度顯示(自動刷新)
  • 進度條 顯示下載速度和預估時間
  • 並行下載 - 同時下載多個模型(可設定 1-5 個)
  • 自動重試 - 最多 5 次,CDN 錯誤自動延長等待時間
  • 重試失敗 按鈕 - 一鍵重新排入所有失敗的下載
  • 自動保存中繼資料 - 模型旁自動保存 .json(含發布日期、觸發詞、SHA256)
  • 自動保存預覽 - 模型旁自動保存 .preview.png.preview.mp4
  • 背景處理 - 瀏覽其他分頁時下載繼續進行
  • 清除已完成 - 移除已完成/失敗的項目

設定

Settings > CivitAI Manager Plus 中設定:

設定 預設值 說明
CivitAI API Key (空) 個人 CivitAI API 金鑰,部分下載需要認證
Auto-organize Downloads 開啟 基底模型 / 作者 / 模型名稱 自動整理
Max Concurrent Downloads 2 並行下載上限(1-5)
Use Aria2 開啟 使用 Aria2 加速下載(不可用時自動退回 HTTP)
Show NSFW by Default 關閉 預設是否顯示 NSFW 內容
HTTP Proxy (空) HTTP 代理 URL(例如 http://127.0.0.1:7890

安裝

方法一:從 URL 安裝(推薦)

  1. 開啟 SD WebUI 的 Extensions 分頁
  2. 前往 Install from URL
  3. 貼上:https://github.com/acer1204/SD-CivitAI-Manager-Plus.git
  4. 點擊 Install
  5. 重啟 SD WebUI

方法二:手動安裝

cd stable-diffusion-webui/extensions
git clone https://github.com/acer1204/SD-CivitAI-Manager-Plus.git

重啟 SD WebUI。


使用方式

  1. 在 WebUI 中切換到 CivitAI Manager Plus 分頁
  2. 輸入搜尋關鍵字並點擊 Search(或按 Enter)
  3. 使用 Filters 摺疊面板縮小搜尋範圍
  4. 點擊模型卡片查看詳細資訊彈窗
  5. 批量下載:勾選模型卡片,然後點擊 Download Selected
  6. 勾選 Save Local Info 可將模型資訊離線快取
  7. 前往 Downloads 分頁查看下載進度

系統需求

  • Stable Diffusion WebUI(AUTOMATIC1111 或相容版本)
  • Python 3.10+
  • requests 函式庫(自動安裝)
  • 可選:aria2c 加速下載
  • 可選:send2trash 安全刪除檔案

授權

This project is licensed under the AGPL-3.0 License.

About

As a frequent CivitAI user, I built this after trying many existing tools — a model manager that's smooth to use and actually stable.

Topics

Resources

License

Stars

Watchers

Forks

Packages