Posts

Yeah our government tends to overreach and are usually SUPER cautious with messing in the first amendment and itd get thrown out, the argument would be if one has the power to do that to ai images then the same power would be given to everyone AGAINST everyone. Most legislation like this is usually overly broad, thereby making pretty much this entire website against those rules as well and all “show accurate” images would have to be removed as that is the artstyle of the show and hasbro would own it. Remember, when laws like this are made they pretty much NEVER protect a single class, as you would have to have laws that define such things beforehand.

I see the argument of recreating artstyles, but at the same time do you want your image to be taken down because you put a happy little tree in the corner of a painting because the estate of bob ross owns that style? The obvious take is people assume “only against ai”, but then youd have to define ai as an entity, that would have to have its own rights, define that, then you can go against it. Right now I know of some cases that have already been thrown out or possibly going to be thrown out because it cannot be proven that ai can recreate an artstyle precisely enough to be copyright infringing as all the example images shown in court (forgot the case, its recent) were shown to be fair use. I say as annoying as it is, is to let the courts sus out ai, our court system is slow and ai is fast so this will take time.

@Ciaran
Yeah I dont upload too many, it CAN be several in a day but then I dont post more for like a few weeks or months while I do other things and or wait for new models or new techniques to come out. 2 a day is fine I guess, 4 for me would work better as thats twitter’s picture limit per post so if I had 4 really nice pics I wanted to put into a twitter post, only 2 of those would get posted here but again, I have like 20 other models that arent pony related and always test things so it isnt too bad of a deal really.

Also ive seen some nice purplesmart pics and novelai stuff but everything I do is offline only models and techniques. Nothing I generate is made using online models so its always more finnicky but way more malleable.

My final word is I think current thing is fine, would be a bummer for me if some of the ones I posted got taken down as I try my best to upload only the ones with no glaring flaws best I can, even if its ALMOST perfect but has a glaring hard to fix flaw (I hate those).

Only thing im confused about is the prompter: and generator: tags, promtper: I guess is me? like prompter:marusame and generator: well that’s vague, there are ai webui’s that can feed multiple models into an image, and even into specific PARTS of an image, like using one model and prompts for the sky, but using a completely different prompt and models for a character and its composited in like comfyui which is node based. So is generator the webui? the model? what about models and LoRa’s? embeddings? For example https://www.derpibooru.org/images/3114109 this image used Pony diffusion, in Automatic1111 but also used a Twilight Sparkle LoRa with a pony specific esrgan upscaler and sometimes I add other VAE’s on top of it.

I guess my point is it can get VERY layered. At the moment ive just been tagging the models I use, but If I wanted I could use a custom anything5 model, with a pony lora, using a random embedding with an anime VAE and itd work, wouldnt even have to use a pony model to get pony pics.

And hands are always an issue, suppose you can do inpainting with an inpainting model to fix it and that would equal editing work and give it more thought, so it works out I suppose. Then you can go a step further and composite multiple generated images together from different models.

Biggest issue is probably pacing, as it seems to be exponential, this will all be irrelevant in 2 weeks with models being able to just not do bad hands or even now you can generate video that keeps the target look locked in.

For even all the images I uploaded that had minor errors, there was easily 100+ that didn’t make the cut, I do try my absolute best to cherry pick only ones with no major flaws and try to get better in the work to not have those flaws in the first place.

And as for the “samey” look of ai art, thats mostly because there are only 2 5 main proper stable diffusion models for pony’s at least on civitai, and they like to “fight” with certain lora’s and artstyles as it wasnt trained for it but thats fixed with more mixing and proper training. But I have been experimenting with ponydiffusionv4 and it seems to like some lora’s and need to get out of my comfort zone with it more which is what I was trying to do with my latest posts.

@Joey  
oh thank goodness, thought something was wrong with my account or something.

anyway to get any info on why my latest upload was removed? was it too unaltered or something to be a webm?

I found a tiny amount of footage of the equestria girls rainbow rocks spanish live musical  
link to the twitter video post  
Its smartphone footage, minus points to the recording person for not going horizontal but it still works i guess.

@trashboat  
Ultraman, i dont know this because i watched the show, but because the SNES game scarred me for life.

@Carcass Strife  
this gif would go great with the Reapers from mass effect sounds.

@Carcass Strife  
really its just zooming in a picture, making a few duplicate layers, and moving it subtly around, mine is 3 frames so just the same pic zoomed in and moved slightly 3 times.
 
What sucks is when you add text you have to copy it the same number of times and merge it to each one of the other frames.

i like this thread because it gives me photoshop project ideas.  
 
like with the vibrating gif thing, so i had my hand at it.

@EHAN  

 
what program is used to make this? i see this sort of “textured” gif on kotaku, ive used several gif programs and gimp and photoshop, all are a pain in the ass.

@Carcass Strife  
yeppers, makes sense though, since its a single account and a gamertag is linked to many many things microsoft server side.
 
Also to avoid name changing abuse id wager.
 
Also first change is free i hear, i dont know since i never changed mine, had it since ’05 before the 360 came out, i remember having to do the xbox to xbox 360 migration.

I wish there was a less time consuming way to add a gif ONTU another picture without creating 50 copies of it, putting the gif frames, frame by frame in between the copies and merging each one down, not to mention the per frame edits i had to do so thats why i only used 14 frames instead of the like 50 or so it had cause damn, thats a lot of time for something so simple heh.

@SquiggyBomb  
i didn’t write gundam as the model, i put the tag as the general show or gist, cause i don’t remember or know of the names of any of them.  
Also, there are 1000’s of anime’s that copy gundam and many are good its just that unless you watch mech anime alot youd have no way to tell it apart >.>

@trashboat  
 
sunbutt agree’s

i like how the update notification only shows a 1, otherwise id be panicking every time i saw a large number attempting to read all the comments.

 
first post to any of these forums as far as I know, and its this.