Posts

@AIpajac
Hrmm, not in a super straight forward way… The “Recent Uploads” technically is meant to satisfy that need but it does show up in second order, so I guess not ideal.

Easiest would be for the tag you already have verified to be slightly different, like prompter:ai-pajac (and a note on it as to why).

The other way around (having people use a separate tag) would be trickier as most people will just put the tag used where they found it on without looking at the description.

Hello. I wanted to ask if there is a way to separate images associated with my tag (prompter:aipajac) from my profile? I mean those images that I don’t publish. I don’t mind people publishing what I once created and what’s floating around the internet, but I would like to keep a certain order on my profile and only show things that I publish myself?

what is something you don’t like or wish was done different about how generative ai is being used for pony and similar fanart stuff? not like scam ads (which everyone dislikes) but like on the creative attempt side

Maybe I should make a “Tools” post at some point. Anyway, extended the RedRcoket captioner demo to manage batch images and .txt / .csv saving: https://github.com/TheSmilingPony/AutoTagger

This was meant to be in the LoRAs thread, whoops…

@Nocturn
If you can run things locally, the two main ways to do this are with inpainting (drawing which areas to use for the current prompt and only filling in that), and regional prompting (aka “forge couple” in reForge) that splits the image into sections (so you can say “this part of the prompt applies to all the image, this part to the upper half, and this part to the lower left quadrant).

Basically it’s possible to work the system, but it’s not straight forward.

@Background Pony #09DE
Since I’m already here… Raw number of images isn’t really a concern, storage is fairly cheap. Transfer is the main cost factor there; site would have to scale close to derpi levels of user numbers and activity for there to be a “problem”. Which IMO would a happy problem.

@Background Pony #09DE
Excuse me if this is off topic, but why do you think that it would be a pressing financial matter? Or that it had been ‘proven’ that it would be? I’m not sure it even matters, but I’m just curious what proof you’re referring to?

oh my gosh that would be so epic - I really can not wait for that to become available. I’m doing two separate images and editing them together now, simply because I can’t get Character A prompted right without it also affecting Character B.

@Admin
I mean financially. As you proved, that’d be a pressing matter sooner than technical limitations

@Background Pony #09DE
On the technical side, more than every existing and defunct imageboard added together.

How many images Tantabus can handle until file storage becomes a problem?

@Teaspoon
Definitely this - one the things I’ve enjoyed most while playing around with various prompts and generators is getting something even better than what I had imagined.

@Background Pony #A669
Concept exploration. Being able to quickly iterate on high level ideas to get varied results is good, as is that it often surprises you with things you hadn’t considered or intended.

i hope this is not too stupid a a question. people who work on animation talk about using what the medium gives you for free. what would you say is the thing that ai gives you for free?

@Teaspoon
Yeah, same here. I’ve had it happen on plenty of non AI art I uploaded as well. I just haven’t really been back to non AI art much since I started doing AI instead.

@tyto4tme4l
@Lord Waite
Yep.

Mind, this isn’t specific to AI, I’ve made loads of vectors where I notice something’s wrong or missing after I considered it “done”, and have heard the same from more “traditional” digital artists.

@tyto4tme4l
Definitely. Happens embarrassingly often when I’m in the middle of tagging something before I hit submit. Often I was distracted by something else in the picture I really liked, or it snuck in while I was changing something else…

@Background Pony #D936
Noticing a glaring flaw once you think you’re finished.

@Teaspoon
what is the most frustrating thing about the process of getting a good ai image?

@Background Pony #D936
Character aware prompting, or some form of smart regional prompting. Ala “Character A has red eyes and yellow hair, character B has blue eyes and red hair, character A is dancing”.

Would just make life easier…

(there’s much more impressive stuff going on ofc)

so what is the most impressive ai development you are excited about?

@Scarlet Ribbon
That’d be nice! I would appreciate it if you’re willing to share with us. I’m always looking for ways to improve my workflow and learn new techniques. Though I’m not in the Discord I’ve seen mentioned so I can only assume that’s where most of that sharing happens.

I did have a question about that. I’m not an artist, but I did wonder if it’s valid for me to generate images to trace. It’d be a step up for me to be tracing something instead of not drawing at all.

@Star Streaker
I’m rebuilding it right now, but this new version is designed to potentially be distributed and is less dependent on Wildcards in favor of ComfyUI nodes that handle everything in the workflow directly. The only things that would be needed would be the various LoRAs that are invoked, and I’m making a toggle for each one of those, with clearly labelled notes that link to the needed LoRA downloads.

@Scarlet Ribbon
That just makes me want to see your workflow even more… Mine are pretty basic, all things considered. I let SwarmUI handle the wildcard stuff.

@Background Pony #D936
Difficult to say. AI-generated backgrounds could be detected, but using AI-generated images as a reference or a base for tracing can’t really be detected, as far as I know.
More popular artists wouldn’t even need a LoRA for their style. Models like ZoinksNoob can already mimic a lot of artists’ styles to some extent.

It would be especially tempting if a lora for your art style already existed, I’d imagine.

(And I’m sure I saw a comment once somewhere where an artist had decided to go all in on AI art and personally trained a lora on their own art style. Not a pony artist, though.)