A collection of enhanced nodes for ComfyUI with metadata handling for Civitai and streamlined model loading.
Added Image Loader for batch load images for Img2Img creations. To use it:
- Create a new folder in the output folder
- Insert a batch of images to the created folder
- Browse and choose output's subfolders from a dropdown list of the node
- All images in the folder will be processed in Img2Img creations
👉 LoRA Metadata Exporter – Extracts trigger words from LoRA metadata (local & online).
👉 Prompt Formatters – Organizes quality tags, descriptions, and prompt structure.
👉 Smart Model Loader – Loads models and LoRAs with automatic metadata extraction.
👉 Enhanced KSampler – Improved metadata tracking with LoRA and checkpoint integration.
👉 Diffusion Model Loader – Like SDXL/PONY version, it loads model along with clips and LoRAs, extracting metadata and and passing them as string.
👉 FLUX Sampler – Metadata tracking with LoRA and checkpoint integration.
👉 FLUX Guidance – The very same FLUX Guidance as ComfyUI original, but with guidance extraction to save as metadata.
👉 Advanced Image Saver – Saves images with embedded metadata, filenames, and auto-numbering.
👉 Custom Latent Generator – Predefined resolutions for common aspect ratios.
👉 Prompt Saver – Don't lose your prompts ever again, by saving their copy in the output folder (or specified subfolder).
👉 Prompt Loader – Load saved prompts easily from the dropdown list, and connect to the Positive/Negative nodes.
Known issue: Occasionally, not every LoRA will be visible on Civitai as the hash will not be recognized correctly. It's most likely caused by models' updates on Civitai, and re-downloading the model solves the issue.
- ComfyUI installed and running.
- Python dependencies (install via terminal):
pip install safetensors
Find Mango Node Pack in the ComfyUI Manager!
OR
- Clone the repository:
git clone https://github.com/mang01010/MangoNodePack.git
- Move the
MangoNodePackfolder into ComfyUI’scustom_nodesdirectory. - Restart ComfyUI to load the new nodes.
💡 Category: Metadata
- The pack started with this node. It extracts trigger words from LoRAs and concatenates them as comma delimited string that can be pushed further, so you won't have to think about these. Don't fail using downloaded LoRA ever again!
💡 Category: Metadata
- This node organizes prompts into three separate text boxes, making them easier to read and edit. It also supports direct integration with the Trigger Exporter.
💡 Category: Metadata
- Does the same thing as Prompt (Mango), but divides prompt further to add embeddings in seperate field.
💡 Category: Loaders
- Loads Stable Diffusion models and up to 5 LoRAs.
- Computes model hashes and applies LoRAs with weight control.
💡 Category: Sampling
- This version of the KSampler node features eight inputs. The standard inputs—model, latent image, positive, and negative—are used for image generation. Additionally, it includes a LoRA stack, checkpoint name, and separate positive and negative text fields as a metadata workaround for:
💡 Category: Metadata
- Saves images with embedded metadata (prompt, seed, CFG, steps, sampler, checkpoint and LoRAs info), ensuring compatibility with Civitai for direct metadata reading.
- Supports auto-numbering and custom subdirectories.
💡 Category: Loaders
- Generates latent tensors in predefined aspect ratios (e.g., 1:1, 16:9, etc.).
💡 Category: Metadata
- Save a copy of prompts in the default output folder (or specified subfolder).
- Load saved prompts easily from the dropdown list, and connect to the Positive/Negative nodes.
Once installed, find the MangoNodePack nodes inside ComfyUI.
Example Workflow:
This project is licensed under the MIT License – see the LICENSE file for details.