comfyui-anima-toolkit

TK Toolkit

简体中文 English

A ComfyUI custom-node pack and local toolbox for LoRAs, prompts, reference images and generated outputs. Built for Anima workflows, with other model families supported by individual tools. It manages inference workflows; it does not train LoRAs.

Current release: 2.37.0. By 时运Toki. The combined distribution including MOSS Layout Studio is GPL-3.0; original MIT grants remain valid. See licensing. QQ community: 758086740.

What’s new in 2.37.0

Restart ComfyUI once after this upgrade, then hard-refresh the page. Later connection changes and model switches take effect on the next request without restarting. Personal credentials and models are not included. Changelog.

Language coverage

Choose Simplified Chinese or English in the toolbox Settings. Version 2.36.0 covers the toolbox, workflow-node interfaces, the gallery workspace and major dialogs. Save your workflow and refresh the ComfyUI page after switching; these frontend changes require no host restart. Prompts, tags, custom names and external service responses stay as written. See node language coverage.

Install

Search for TK Toolkit in ComfyUI-Manager. Its Registry ID is anima-toolkit (Registry page).

Or clone it into your ComfyUI installation:

cd ComfyUI/custom_nodes
git clone https://github.com/Ararararararaki/comfyui-anima-toolkit ComfyUI-Anima-Batch-LoRA

Restart ComfyUI. If dependencies are missing, install requirements.txt with the Python environment that runs ComfyUI. Optional dependencies are in requirements-optional.txt.

The built app/ toolbox is included. Normal use requires neither Node.js nor a frontend build.

Open the toolbox and choose English

Click TK Toolkit / 本地工具箱 in ComfyUI’s toolbar, or the panel button in the batch LoRA loader. Search for TK in the node menu to find the workflow nodes.

For the folder above, the direct entry is the local toolbox. Use your actual ComfyUI address/port if different. Chrome or Edge on localhost provides the fullest folder-access support.

Open Settings (设置) → Language / 语言 → English. The choice is saved in this browser and the toolbox reloads. Switch back to 简体中文 from the same control.

Some workflow-node labels/tooltips, historical changelog entries and detailed documentation remain in Chinese. The language control does not translate your prompts, model names, paths or notes.

Core features

Task Tools
LoRAs Batch loading, individual weights and trigger words; recursive local scans, subfolder categories, Civitai matching and downloads
Prompts Prompt Library and Prompt Cards, bilingual tags, translation, normalization and autocomplete
Image analysis Read saved prompts, workflows and generation parameters from PNG, JPEG and WebP metadata
Outputs Multiple folders, work/character browsing, search, favorites, ratings, preview edits and export metadata controls
References Multi Gallery: Danbooru, Civitai, Pixiv, Safebooru, yande.re and Konachan.net
Workflow helpers TK Prompt Agent for Anima / Krea 2, sequential prompt batches, camera/pose tools and text/image routing

Prompt Agent requires your own compatible model API connection. Models, personal credentials and derived recognition caches are not bundled. See the English feature summary, full Chinese manual and Prompt Agent guide.

Pixiv receiver bridge

The optional TK → Pixiv receiver bridge 0.2.0 transfers an Outputs image or edited preview into Pixiv’s image upload control.

  1. Download the archive from Updates & QQ group in the toolbox, or obtain it from integrations/pixiv-drop-bridge. Extract it into a permanent folder.
  2. Open chrome://extensions/ or edge://extensions/. Enable Developer mode, choose Load unpacked, and select the folder directly containing manifest.json. This first installation is manual.
  3. Use the same browser profile for the toolbox and Pixiv upload page, then refresh both pages. After an extension update, reload the extension and refresh both pages again.
  4. Hover over an Outputs image until its copy is ready, or use Drag current image in the preview. Drag it onto Pixiv’s upload area or the bridge box. If cross-tab dropping is missed, drag once in the toolbox and use the bridge’s receive button on Pixiv.
  5. Check that Pixiv itself shows the image preview. You review and submit the post yourself.

The local toolbox port is detected automatically; no address, port or API key is entered. Supported scope: Chrome/Edge, local HTTP localhost / 127.0.0.1, one image at a time, up to 32 MiB. Remote/LAN addresses, HTTPS toolboxes and other browsers are outside this scope. Exported bytes follow the toolbox’s metadata and attribution settings. The bridge guide records testing limits and troubleshooting.

Update and keep your data

Use the loader’s Update (更新) button or ComfyUI-Manager. For Git installations, run git pull inside the existing plugin folder. Finish active generation work, restart ComfyUI through your usual launcher, then hard-refresh with Ctrl+Shift+R.

The built-in updater preserves data/, models, input/output folders, credentials and user configuration. The changelog contains the full release history in Chinese.

Browser storage is separate. Keep the same browser profile and address/port: localhost, 127.0.0.1 and different ports are different browser sites. Language, appearance and browser databases belong to that site. Export outfit cards with the Outfit Library’s own Export control before clearing site data or switching browsers, and retain the plugin’s data/ folder. Folder permissions may need to be granted again after moving or importing data.

Repository · Issues · Machine-readable summary