Home Creators Posts Import Register Favorites Logout

Content

Test Environment:win11 / RTX3060 / comfyui 0.27.1

The generation logic of the NO8D-Generate node has been further optimized, resulting in a drastically faster generation speed as the most obvious improvement. Generating one FHD image previously took 25 seconds, which has now been reduced to 17 seconds, representing a roughly 40% speed boost.

The NO8D-A/B preview node has added a function to display image dimensions

Fixed the bug where the NO8D-Prompt-view node outputs nothing when running independently

The output logic of the NO8D-Load-images node has been optimized. Double-click an image to view its enlarged version. Single or multiple selected images can be exported at once; all images will be exported if none are selected. An eye icon is added on each image, and images with the eye icon turned off will not be exported.

The NO8D save node has added a metadata interface to reshape the image metadata and rewrite it into your desired content.

The API has added support for doubao-seed. Tests show its capabilities of prompt expansion and image reverse inference are significantly superior to the Qwen series.

Minor tweaks and fixes will not be elaborated on. The current version runs fast and stably, nearly matching the state I envisioned. Barring unforeseen circumstances, there will be no frequent updates going forward unless I come up with new optimization ideas.

The new version has been submitted to ComfyUI. You can download it directly from GitHub if you want to use it immediately.

https://github.com/no8d/ComfyUI-NO8D-controls

OK, the above is all the updates for this time. Now you can directly obtain "NO8D-Controls" by searching for it in comfyui-manager. "NO8D-Controls" is entirely designed to improve efficiency and enhance control. I hope you will like it.

🫡 May the AI-power be with you, see you soon !🫡

Disclaimer: My model or workflow is solely for learning and communication purposes and shall not be used for any commercial activities. All responsibilities for the generated content, including its legality, ethical implications, or potential misuse, are borne entirely by the user. By using this model or workflow, you agree to this principle. Please use it responsibly within the scope of local laws.

Files

Comments

Wylmquest

"all images will be exported if none are selected...." this is particularly annoying because inference starts immediately and the system is overloaded when you have many references. It should be the opposite. When Nothing is selected there should no action, just notification to select something. As pictures get de-selected in some situations, this node created issues for me several times by trying to load all references ... Suggest you change that in next update.

NO8D's AI-Workshop

Oh, I'm terribly sorry for the trouble this has caused you. Admittedly, everyone has different usage habits, but your suggestion is quite reasonable, and I will adjust this usage strategy as soon as possible.

Wylmquest

Also, the a/b compare node degrades resolution of the previews. other compare nodes in comyui don't do that. Check your own against e.g. image comparer by rgthree.

Wylmquest

from the generating node you removed latent output. why? I find the latent output useful because it avoids in some situations VAE decoding and encoding.

Wylmquest

Thank you for your response. The lora stacker allows me to see loras in one particular folder called Flux2. I have many other lora folders. Is there a way to use them?

Wylmquest

Sorry for all my questions. How do I transfer collection of references from old node to the new node? I don't want to recreate this collection one by one. That would be a waist of time.

Wylmquest

I am also very disappointed your new workflow is a pined-up bowl of noodles. I explained to you on Discord how to make correct workflows, maybe you had no time to read my comments?

NO8D's AI-Workshop

If you're referring to the node layout, I apologize. As I mentioned before, everyone has different work habits, and node layout varies from person to person.

NO8D's AI-Workshop

There will be no latent after local redrawing. If you need latent, you can add a VAE encoder after the image output.

NO8D's AI-Workshop

The LoRa Stacker node scans ComfyUI's default loras folder; theoretically, any LoRa stored within this folder can be loaded.

Wylmquest

i mean that i organize a collection of references from many different locations and then the node is full links. as soon as i change the node all the links disappear and i have to start from beginning. There should be a way to transfer content of this node to a newer nod. I am afraid you will publish newer versions frequently and this mins lost work for me.

Wylmquest

you do not need to apologies. you do a grate job here. I just want you to adopt one good standard. Open flowchart with logical flow. I don't care, and you should not either, which bad habits other people have. Use good standards and other will follow.