they even nerfed gpt4
they even nerfed gpt4
Fso horrificly
Fcant even handle over 4k
Fcontext lenght recently
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=40)
Dsome have said its a mixture of real video footage
Dhuge library from shutterstock
VIsn't there any way we can unnerf these models?
VI have seen people talk about uncensored ai models..and removing all the safety
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=16)
VYou could be right .they can do that to wow all the people and raise money from investors....
VMake extreme good videos...then raise money lots of it then when the final version comes say that it has been heavily nerfed so that people won't misuse ..or say that it requires really powerful system for it to deliver great quality videos...
Fye they raise money
Fnewest batch of sora videos arent so great
FPrompt 3: A young professional product reviewer in a well lit video studio is surrounded by gadgets and technology, sitting in front of a computer with two displays. He's holding a cinema camera as he ponders what video to make next. He is in focus, while the background is…
2/29/24, 10:39 PM
Fwhat they don't said is how many inference time need those 1 minute videos XD
Fjust the same slow moving melted crap you always see with AI videos
FGenerating.... please come back on saturday....
Fdog one was ok, but feels much lower quality than the ones from 2 weeks ago
Feven gpt4 became so slow
VAnd dumb..even the copilot as well..sometimes I tell it do something specific but instead of doing what I tell it to do .it starts expanding on what I tell him without exactly doing what I tell it to .then in 2nd prompt I have to tell it not to elaborate or expand ..just do what you are said .then it does it.
AAnyone talking shit about Sora must surely realize by now that nearly everything they can think of that can be criticized has probably already been criticized to some extent by the Red Team, and their internal model(s) are likely so much better still.
AWhen sharing screenshots is it unsafe to share ports that display in things like llava?
VWhat they release to the public matters because NORMAL people like us will only get heavily nerfed AI with hindered capabilities........
Agotta be a special kind of twisted to enforce a cultural divide right down to the accessibility of aesthetic content
Awhat they’ll probably do is find a way to label it all as “computer-mediated semantic collages” which would then allow them to distance it from art and control it even further.. which will really suck
Aand then a war on semantic collage becomes a proxy war for art and culture, history repeats, etc. okay I’m done depressing discord, sorry
Rwonderful article, I loved it. Thank you specially for the ComfyUI screenshot, I got it going in about 5 seconds. It is a very impressive model indeed!
R@Dr. Furkan Gözükara accidentally closed the live Supir gradio app. There's 1000 images running. How do i find the link again?
RMaybe for some later update, might be worth keeping that link on each image upscale print??
RAnd perhaps, also, the image count out of total, ie 455/1000, so we can see where we are at in the process
Fi think auto1111 already has this
Fyou can see terminal
Fit shows
Fthe link is also at the beginning of terminal
Fyou can also see from your browser history
Fbest option
RYes but after 100 images you lose the beginning
Ftrue
Fso best is browser history
RBrowse history
RTrue
Ractually oddly, when i go to e7ab2bbe0a184b3859.gradio.live there's no tracking of time and image count. It does connect to gradio but no information is displayed, even though i can see from terminal it is still going the process
D
! I cant find it in A1111, this is not the same as having one main and one style, its taking both as main. As an example 3 positive main and one negativeFye gradio will get broken
DI just use the output folder to view the images if i am having a large batch
Fit has BREAK keyword
FGitHub
Stable Diffusion web UI. Contribute to AUTOMATIC1111/stable-diffusion-webui development by creating an account on GitHub.
FGitHub
Stable Diffusion web UI. Contribute to AUTOMATIC1111/stable-diffusion-webui development by creating an account on GitHub.
DThere is a better way to do it, i dont use gradio to open the URL i use the IP Address, through a vpn which provides static IP for each device. “Tailscale” hope it resolves your issue 
DLet me know if it works for you. Note both devices need to have the app