[Stable Diffusion] Prompt Sharing and Learning Thread

[moc587]

since my initial post it loaded. come to find out its just taking a long time to load. over an hour. ugh!. no idea what it is. didnt happen on Win 10. i will try what u suggested.

wat does it say in the terminal any errors? try deleting the "venv" folder and launch it again. it usually recreates it when you launch it again. but if it didnt you might have to manually recreate it.

 

[modine2021]

wat does it say in the terminal any errors? try deleting the "venv" folder and launch it again. it usually recreates it when you launch it again. but if it didnt you might have to manually recreate it.

it shows no errors at all.

 

[JhonLui]

it shows no errors at all.

Just a assumption, but since you just went to W11, yo might have all the nice toys (Recall, bitlocker, copilot..) activated.
Specially Recall is infamous for slowing down or even freezing systems due to its tasks. You might want to check it out, or just take a look at the ram/cpu usage.
Also I hope you made a clean install and debloated your system

 

[Synalon]

since my initial post it loaded. come to find out its just taking a long time to load. over an hour. ugh!. no idea what it is. didnt happen on Win 10. i will try what u suggested.

I'm pretty sure WIN 11 has some background apps that are resource heavy, you can try disabling those and running Comfy/Framepack etc to see if it helps, just keep track of what you disable in case you want to turn them back on again later.

You must be registered to see the links

There are a few listed here, the first one helps to run old games so I would leave that on myself.

 

[moc587]

it shows no errors at all.

hmm well at this point i'd just do a full reinstall. less headache that way. move the models folder somewhere safe and keep a note of all the custom nodes you have in custom_nodes folder. i personally use

You must be registered to see the links

method. i use linux so manual install is my only option. i use miniconda for the virtual env. so whenever i break my setup with some random bullshit custom nodepack i install, i just nuke that virtual env, recreate it and "pip install -r requirements.txt" and for custom nodes, do it with comfyui manager, it doesn't take that long. i prefer having more control but if you dont wanna do all of that use the portable package from github releases not the desktop application. i dont like the new app thing they have that bundles its own browser and stuff.

 

[oldmasterxd]

just drop to share 1020e1242302304320x than any ai make ur games animations etc work wonders this is one exemple but for games its not good keep realistic add render 3d or cartoon whatever realistic we have real porn already

View attachment download (9).mp4

where can i find Lora that are banned from civai? i need a bestiality lora for a rpg games(mainly wild animals, like wolf, boar, bears... like skyrim)

 

[Austin Powers Shagadelic]

With Civitai removing lots of Loras I know there are archives but where are new Loras being uploaded if they're not welcome anywhere? Also for those few that are only doing paid loras (looking at you tensor.art) would it be something we can share on this forum or is there a Kemono.party type tool for tensor.art ?

 

[JhonLui]

With Civitai removing lots of Loras I know there are archives but where are new Loras being uploaded if they're not welcome anywhere? Also for those few that are only doing paid loras (looking at you tensor.art) would it be something we can share on this forum or is there a Kemono.party type tool for tensor.art ?

I haven't got any official news, but the comment I red in this model yesterday is disturbing:

You must be registered to see the links

If Civitai is going to ban Nsfw, not only will make them shut down as it represents 90% of the content, but it'll also lead to a small "dark age" in contents before new sources surface.

Therefor yes, it's due to start at least gathering all the informations about sharing sites/torrents along the way and forming a list to pin on the O.P.

Let the hunt begin..

 

[Sepheyer]

I saw folks are posting LORAs on HuggingFace. This is especially true for Qwen - bunch of those went up on HF, but never showed up at Civitai. May be true for others as well, but I mostly pay attention to things-Qwen.

 

[Synalon]

I haven't got any official news, but the comment I red in this model yesterday is disturbing:

You must be registered to see the links

If Civitai is going to ban Nsfw, not only will make them shut down as it represents 90% of the content, but it'll also lead to a small "dark age" in contents before new sources surface.

Therefor yes, it's due to start at least gathering all the informations about sharing sites/torrents along the way and forming a list to pin on the O.P.

Let the hunt begin..

From what I understand Civtai wants NSFW images to have metadata now to be visible, they are trying to moderate using some shitty ai botthing that mistakes cum for vomit and other dumb things. By having access to the metadata the bot can see your prompt and intentions at least I guess.

I like to keep the prompts secret when possible though, so I don't upload there much.

 

[JhonLui]

From what I understand Civtai wants NSFW images to have metadata now to be visible, they are trying to moderate using some shitty ai botthing that mistakes cum for vomit and other dumb things. By having access to the metadata the bot can see your prompt and intentions at least I guess.

I like to keep the prompts secret when possible though, so I don't upload there much.

Good.. psychopol is getting real..
At least that explains why the search now goes trough metadata giving messy results.

Just to be sure.. you are referring to online generation when you're talking about prompts?

 

[Synalon]

Good.. psychopol is getting real..
At least that explains why the search now goes trough metadata giving messy results.

Just to be sure.. you are referring to online generation when you're talking about prompts?

I don't think it matters if you generated the image locally or online, the metadata should show what prompts you used to generate the image.

For example "blonde hair, blue eyes,curvy figure, big breasts".

If you upload the image after editing it with something like photoshop or just save it as a jpeg it will remove the metadata.

Without that metadata the bot can't "see" what the prompts used were.

Essentially, it means if you upload the image that came out of your image generator without editing it, it should be fine.

Just to clarify though, this is just my understanding of it, don't be surprised if I end up being wrong.

 

[JhonLui]

I don't think it matters if you generated the image locally or online, the metadata should show what prompts you used to generate the image.

For example "blonde hair, blue eyes,curvy figure, big breasts".

If you upload the image after editing itwith something like photoshop or just save it as a jpeg it will remove the metadata.

Without that metadata the bot can't "see" what the prompts used were.

Essentially, it means if you upload the image that came out of your image generator without editing it, it should be fine.

Just to clarify though, this is just my understanding of it, don't be surprised if I end up being wrong.

Ok, I tought for a second that they can send a bot to your local installation...
Obviously if you upload the image in bmp or other than jpg it will always contain metadata, it was so even long before AI.
You can still select the option not to include them, but they can be extrapolated anyways in a general manner.

 

[Synalon]

Ok, I tought for a second that they can send a bot to your local installation...
Obviously if you upload the image in bmp or other than jpg it will always contain metadata, it was so even long before AI.
You can still select the option not to include them, but they can be extrapolated anyways in a general manner.

It is only images you upload to them that the bot can look at the metadata for, anything that isn't on their website they can't check.