Sick of unwanted <pad> tokens disrupting your prompts? This tool provides a reliable solution by automatically cleaning up caption text and saving it into a neatly formatted .txt file, enhancing your workflow in ComfyUI.
- Strips unwanted tokens from text, ensuring cleaner and more accurate prompts.
- Implements strict validation for file paths and names, enhancing security and preventing errors.
- Integrates seamlessly with captioning nodes like Florence and supports LoRA-style datasets.
Context
ComfyUI-E-Tier-TextSaver is a specialized node designed to enhance the text processing capabilities within ComfyUI. Its primary function is to remove extraneous tokens, such as <pad>, from caption text generated by models like Florence2, thereby producing clean text files that are essential for various workflows.
Key Features & Benefits
This tool effectively eliminates unwanted tokens, which can clutter prompts and potentially lead to errors in AI model training. The strict path validation ensures that users can only save files to existing directories, mitigating risks associated with invalid file paths and enhancing overall reliability.
Advanced Functionalities
The node features a customizable save_to parameter, which allows users to define specific output paths through a YAML whitelist. This flexibility enables users to select from predefined directories, making it easier to organize and manage saved text files according to their project needs.
Practical Benefits
By streamlining the process of saving clean caption text, this tool significantly improves the efficiency of workflows in ComfyUI. It allows users to focus on their creative tasks without worrying about the integrity of their text files, thus enhancing control over the data used in AI model training.
Credits/Acknowledgments
This tool was developed by E-tier-newbie and NDY, contributing to the open-source community and offering valuable resources for ComfyUI users.