ComfyUI-Multi-Lora

English | 中文

ComfyUI-Workflow-Tools

A collection of custom nodes for ComfyUI.

Installation

Clone into your ComfyUI custom_nodes folder:

cd ComfyUI/custom_nodes
git clone https://github.com/Danchi93/ComfyUI-Workflow-Tools

Restart ComfyUI.

Language

The UI defaults to English. To switch to Chinese, change const LANG = "en" to "zh" at the top of each .js file in the js/ folder.

---

Multi LoRA Loader

Manage multiple LoRAs in a single node.

Features

• Enable/disable each LoRA with a toggle switch
• Adjust weight per LoRA (click or drag)
• Custom trigger words per LoRA with local auto-memory
• Triggers output — concatenates all enabled LoRA trigger words into a single STRING output
• Optional CLIP input/output — connect a CLIP model to also apply text-encoder weights of LoRAs; entirely optional (skip it for model-only LoRAs)
• URL button per LoRA — save a download page link (Civitai, LiblibAI, etc.) and jump to it with one click
• Thumbnail previews — place an image next to your LoRA file (lora.png or lora.preview.png) to see a thumbnail in the dropdown; hover to view it larger
• Add notes and remove LoRAs dynamically

Usage

Add the Multi LoRA Loader node, connect the model input, then use the model output as usual. Connect the triggers output to any text input node (e.g. Prompt Segments) to automatically inject trigger words.

URL button

• Grey ＋ — not configured. Click to paste the download page link (saved automatically and remembered per LoRA).
• Green ↗ — configured. Click to open the link. Hold Shift + click to edit the URL.
• Trigger words and URLs are remembered in your browser (localStorage), so you only set them up once — they auto-fill whenever you select that LoRA again.

Thumbnails

Put an image with the same name as your LoRA in the same folder:

models/loras/
  style.safetensors
  style.png            ← shows a small thumbnail next to the name
  style.preview.png    ← also supported

Hover a row to see a large preview next to the dropdown.

---

Prompt Segments

Combine multiple prompt segments into one node with auto-complete, ordered output, and direct CLIP conditioning.

Features

• Add/remove prompt segments dynamically
• Enable/disable each segment individually with a toggle
• Label each segment for quick organization
• Tag auto-complete — type 2+ characters to see suggestions from a 10,000+ Danbooru tag dictionary; click a tag to insert it, or use arrow keys + Tab to navigate with keyboard
• Insert position control — choose where the incoming prompts_in text is placed (before segment 1, between segments, etc.)
• Direct CLIP output — optional CLIP input/output lets you connect straight to the KSampler without a separate CLIP Text Encoder node
• Outputs all enabled segments merged in order

Usage

Add the Prompt Segments node. Connect prompts_out to a CLIP Text Encoder (or any text input), or optionally connect a CLIP model to use the conditioning output directly on a KSampler.

---

Resolution Switcher

Quickly switch between pre-configured resolutions without manually re-typing dimensions every time. Outputs a blank latent with the selected size.

Features

• Pre-define multiple resolution presets (Width, Height, and Batch Size)
• Switch between resolutions instantly using toggle switches
• Dynamically add or remove resolution presets
• Eliminates the need to repeatedly input dimensions when testing different aspect ratios

Usage

Add the Resolution Switcher node. Pre-set your frequently used dimensions, toggle the one you want, and connect the Latent output to your KSampler or any node that accepts latent input.