ComfyUI FE 1.33.9: PNG Text Data Missing

by Admin 41 views
ComfyUI FE 1.33.9 Text Issue: A Comprehensive Analysis

Hey everyone, let's dive into a frustrating issue some of us are experiencing with ComfyUI FE 1.33.9. It seems like there's a problem where text data isn't being correctly pulled from .png files when you drag and drop them onto the canvas. This is a real bummer, especially if you're like me and rely heavily on reusing prompts and workflows. I'm going to break down the problem, what we've tried, and how to potentially fix this issue. We'll explore the expected behavior, the actual behavior, and steps to reproduce the problem. Let's troubleshoot this together, so we can get back to creating awesome AI art!

The Core of the Issue: Text Data Disappearing

The central problem, guys, is that after updating to FE 1.33.9, when we drag and drop a .png file (that contains a ComfyUI workflow) onto the canvas, the workflow loads up, but the text prompts and any text in Notepad nodes are empty. This means the text data isn't being correctly read from the .png. It's like the software is forgetting the text instructions. We're talking about the CLIP Text Encode nodes, and any other nodes where you've input text. This is a pretty significant setback because it forces us to manually re-enter our prompts, which is time-consuming and reduces the efficiency of our workflow. The issue seems to be consistent across different models and workflows, making it a widespread problem. This issue has occurred in the past but wasn't reported. Let's analyze how to fix this situation.

This behavior is not ideal. It breaks the workflow reuse capability. It would be nice to load a .png and immediately start generating images based on the saved prompt.

Expected vs. Actual Behavior: A Clear Discrepancy

Let's get into the nitty-gritty and analyze what should happen versus what actually happens. The expected behavior is that when you drag a .png file onto the ComfyUI canvas, the workflow, including all its nodes and their text data, should load up perfectly. The prompts in the CLIP Text Encode nodes, the text in the Notepad nodes - everything should be exactly as it was when the .png was saved. This would let us pick up right where we left off, tweaking parameters, and generating new images without manually retyping everything.

However, the actual behavior is different. We drag the .png file, the workflow loads, but the crucial text fields are empty. This is the new Flux2 Template that Comfy provided and it is not modified. That's a significant difference that breaks the workflow and increases our workload. The issue goes beyond a single workflow; it affects many different .png files from different models and various workflows. This suggests a problem with how the latest ComfyUI version reads the text data embedded within the .png files.

Steps to Reproduce the ComfyUI Text Issue

If you're experiencing this issue, it's essential to understand how to reproduce it so you can verify if a fix has worked. Here's a clear guide on the steps to reproduce the text issue:

  1. Update to FE 1.33.9: Make sure your ComfyUI is updated to the latest version, FE 1.33.9. This is the version where the problem has been confirmed.
  2. Locate a .png File: Find a .png file that has a saved ComfyUI workflow. This .png should contain text within its nodes.
  3. Drag and Drop: Drag and drop the .png file onto the ComfyUI canvas. Wait for the workflow to load.
  4. Observe the Text Fields: After the workflow has loaded, check the text fields. Specifically, look at the CLIP Text Encode nodes and any Notepad nodes. They should contain the text prompts and notes from the original workflow.
  5. Verify the Problem: If the text fields are empty, then you've successfully reproduced the issue. The .png file has loaded the workflow, but the text is missing, which confirms the problem.

By following these steps, you can reliably test whether the text data in your .png files is being correctly loaded in ComfyUI FE 1.33.9.

Troubleshooting and Potential Solutions

Let's explore some troubleshooting steps and potential solutions to resolve the text issue:

  1. Disable Custom Nodes: Try disabling all custom nodes. Custom nodes can sometimes interfere with ComfyUI's core functions. Follow the instructions to disable all custom nodes. Restart ComfyUI and then test the issue. If the text data loads correctly with custom nodes disabled, then the problem lies with one or more of your custom nodes. You can re-enable the custom nodes one by one to identify the culprit.
  2. Reinstall ComfyUI: If disabling custom nodes doesn't work, consider reinstalling ComfyUI. A fresh install can resolve corrupted files or configuration issues that might be causing the problem.
  3. Check for Updates: Make sure your ComfyUI, as well as any dependencies, are up to date. Outdated components can create compatibility problems. Also, ensure your system meets the minimum requirements.
  4. Examine the Debug Logs: Carefully review the debug logs for any error messages related to loading the .png file or processing text data. The logs might point to the root cause of the problem. Also, verify that the CLI logs are not showing any errors when loading the .png file.
  5. Community Forums and GitHub: Visit the ComfyUI community forums or the GitHub repository to see if others are experiencing the same problem. There might be a temporary fix or workaround available. Also, report the issue on GitHub if it hasn't already been reported. Providing detailed information, including steps to reproduce the issue, can greatly assist the developers in resolving the issue.
  6. Test Different .png Files: Try loading various .png files. If the problem persists across all files, it's more likely a system-wide issue rather than an issue with a specific workflow.

By following these steps, you can find the root cause of the text issue. Remember to provide the developers with as much information as possible to expedite a solution.

Debug Logs Analysis: A Glimpse into the Problem

The provided debug logs give us some clues, although the CLI didn't show any actions when the .png was dropped onto the canvas. Here's a breakdown of what the logs tell us:

  • Model Loading: The logs show the loading of various models, including Flux2TEModel_ and AutoencoderKL. These models are essential for generating images, and their successful loading suggests that the core system is functioning correctly.
  • Warnings: There's a warning about TAESD previews, but this is unlikely to be the cause of the text issue.
  • Prompt Execution: The logs indicate the execution of prompts and the time it took to generate images. This confirms that image generation is still functional, even if the text data isn't being loaded properly.

From these logs, we can infer that the core image generation processes are working fine. However, since there is no log related to loading the text data, we cannot pinpoint the exact cause of the issue from these logs. To get more clarity, we'll need to look at the logs generated when a .png is dropped onto the canvas.

Love for ComfyUI and Future Hopes

Despite this hurdle, I still love ComfyUI! It's an incredible tool for AI art generation, and I'm sure the developers will fix this text issue. Let's keep supporting ComfyUI and working together to improve this fantastic software.

I hope this helps! If you have any additional information or solutions, please share them. The goal is to get this resolved as quickly as possible.