TA-ComfyUI-Nodes-Pack

A collection of various useful ComfyUI custom nodes

Tested with Python 3.12 and pytorch 2.9.1+cu128 and LM-Studio-0.3.34-1-x64

Installation

Requirements

This custom node pack requires:

• ComfyUI-GGUF - Install it via ComfyUI Manager

Installing ComfyUI-GGUF

1. Open ComfyUI Manager
2. Search for "ComfyUI-GGUF"
3. Click "Install"
4. Restart ComfyUI

---

1. Model Loaders

1. ta_load_diffusion_model_with_name
2. ta_load_checkpoint_model_with_name
3. ta_load_gguf_model_with_name

These are model loaders with an additional output that displays the model name so that it can be used in the further workflow.

---

2. Extension for EBU-LMStudio

(https://github.com/burnsbert/ComfyUI-EBU-LMStudio)

1. ta_lmstudio_load_on_run (text2prompt & image2prompt)
   Loads the LLM Model in the LM Studio server when you click RUN
2. ta_lmstudio_model_selector (ext2prompt & image2prompt)
   Selection field for all models installed in LM Studio. Vision models are marked with (V).
3. ta_ebu_lmstudio_vision_node (image2prompt)
   Settings for Image2Prompt Generation

These are extensions for the EBU-LMStudio custom node. The ComfyUI-EBU-LMStudio node must be installed on your system.

---

3. Filename Generator

The TAFilenameGenerator is a utility node for ComfyUI designed to automatically generate complex and organized file paths for saving images.

It outputs two separate strings:
filename: The complete path and name for the main generated image.
filename_up: A corresponding path and name for an upscaled version (it adds a suffix like "UPSCALED").

Key Features
Dynamic Folders: It allows you to specify a main output_folder and a subfolder.

Date Parsing: The subfolder field can parse date/time codes (e.g., an input of %Y-%m-%d will be converted to a folder named 2025-11-13).
Custom Naming: It builds the filename by combining several user-defined widgets:

name_prefix (e.g., "TA")
wf_version (e.g., "v2.50")
The model_name (from its input)
A date_format (selected from a dropdown list)
A delimiter (e.g., "-")

Ready to Use: The final strings (like TA-Outputs\2025-11-13\TA-v2.50-Modelname-202511132027.png) are passed to a SaveImage node, which will automatically create the directories and save the image.

---

4. Prompt Controller

TA Prompt Controller (Switch) A custom node for ComfyUI designed to manage and route text inputs within a workflow. It serves as a central control unit to switch between manual text input and generated text (e.g., from LLMs or Vision models).

Key Features:
- Dual Input: Accepts both a manual text string and a generated prompt string.
- Mode Switching: Offers five modes to control the output:
1. Manual Only: Passes through the user's manual text, ignoring generated input.
2. Generated Only: Passes through the generated text (e.g., from LM Studio), ignoring manual input.
3. Combine: Manual + Generated: Prepends the manual text to the generated text.
4. Combine: Generated + Manual: Appends the manual text to the generated text.
5. Clear / Empty: Outputs an empty string (acts as a reset/mute).
- Custom Delimiter: Allows defining a separator (e.g., , ) for combining modes.

Use Case: Ideal for workflows that alternate between Text-to-Image (manual prompting), Image-to-Prompt (vision analysis), and Text-Expansion (LLM enhancement) without needing to rewire nodes.

---

5. Directory Captioning

The TA Directory Captioning node is a powerful tool for batch-processing images in your workflow.

Functionality:

1. LM Studio Integration: Connects seamlessly to a local LM Studio server, utilizing a user-selected Vision Language Model (VLM) for multimodal processing.
2. Batch Captioning: Takes a specified folder path, iterates through all image files (.png, .jpg, etc.) within it, and sends each image to the VLM.
3. English Output Focus: Generates detailed, descriptive captions specifically tailored for use as prompts in English, ensuring high quality for subsequent text-to-image generation.
4. Automatic Saving: Saves the generated caption for each image as a corresponding .txt file in the same directory, enabling easy integration with datasets and other workflow steps.

---

6. TA Save Image & Prompt TXT

Functionality:

1. The TA Save Image & Prompt TXT node is a node that can save an additional .txt file in your output-directory with the used positive and negative prompt inside. You can switch this functionality on or off.

---

v1.0.16 (2025-12-18)

1. TA Save Image & Prompt TXT Node: I fixed a bug that prevented the workflow data from being saved correctly in the .png file.

---

v1.0.15 (2025-12-18)

1. changed TA Save Image & Prompt TXT Node (it can save now the positive and negative prompt)

---

v1.0.14 (2025-12-17)

1. added TA Save Image & Prompt TXT Node

---

v1.0.13 (2025-12-12)

1. added TA Directory Captioning Node

---

v1.0.10-v1.0.12 (2025-11-25)

1. increased max. possible max_tokens
2. increased max. possible context_lenght
3. increased max. possible wait_time

---

v1.0.9 (2025-11-25)

1. added ta_prompt_controller, a node designed to manage and route text inputs within a workflow.

---

v1.0.6 (2025-11-13)

1. added ta_filename_generator, a node for creating image output filenames and folders

---

v1.0.5 (2025-11-01)

1. added ta_lmstudio_smart_loader, a simple and fast loader for LM Studio Models
2. changed ta_lmstudio_load_on_run for more control if needed