|
LifeSunDeath fucked around with this message at 04:49 on Feb 18, 2024 |
# ? Feb 18, 2024 04:41 |
|
|
# ? May 28, 2024 07:58 |
|
XYZAB posted:I wasn't trying to get anything pornographic, and then here's Natalie Portman standing in a pool of water pouring milk all over herself with no clothes on. It was one generation of like 20 that it spat out and the only one that had any nudity, so maybe it was an abberation, but still. Oh yea screen all images during any demonstrations lol
|
# ? Feb 18, 2024 04:53 |
|
XYZAB posted:I wasn't trying to get anything pornographic, and then here's Natalie Portman standing in a pool of water pouring milk all over herself with no clothes on. It was one generation of like 20 that it spat out and the only one that had any nudity, so maybe it was an abberation, but still. Please share that, so we know what to watch out for. For the children y'know.
|
# ? Feb 18, 2024 19:20 |
|
|
# ? Feb 18, 2024 20:44 |
|
KakerMix fucked around with this message at 22:13 on Feb 18, 2024 |
# ? Feb 18, 2024 21:22 |
|
I'll see you and raise you... an acorn, object of abject mortal terror, the face of imminent and immediate danger of death, excessive use of control to resistance is sustained Steps: 48, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 1, Size: 512x512, Model hash: f65275abcf, Model: notreallyistic_v2, Denoising strength: 0.32, Clip skip: 2, RNG: CPU, Hires upscale: 2, Hires steps: 32, Hires upscaler: 4x-UltraSharp, Version: v1.5.2
|
# ? Feb 18, 2024 22:59 |
|
Hey, guys. Just so you know, AI says that consciousness is stored in the brain balls:
|
# ? Feb 19, 2024 00:28 |
|
https://packaged-media.redd.it/td4k...5a4702295cf#t=0
|
# ? Feb 20, 2024 03:41 |
|
this is actual video of will smith and not AI video
|
# ? Feb 20, 2024 04:06 |
|
TIP posted:this is actual video of will smith and not AI video yes e, more SORA vids: https://v.redd.it/waec54ai4rjc1/DASH_720.mp4 LifeSunDeath fucked around with this message at 18:35 on Feb 20, 2024 |
# ? Feb 20, 2024 04:28 |
|
|
# ? Feb 20, 2024 22:31 |
|
I don't know when this changed, but now Midjourney is only a thousand images before you get access to the web interface.
|
# ? Feb 21, 2024 07:42 |
|
Humbug Scoolbus posted:I don't know when this changed, but now Midjourney is only a thousand images before you get access to the web interface. Why can't I just use it from the get go if I'm paying?
|
# ? Feb 21, 2024 14:25 |
|
Humbug Scoolbus posted:I don't know when this changed, but now Midjourney is only a thousand images before you get access to the web interface. Like you need to generate 1,000 images via Discord before you can do so on their site? Guess I should just start hammering away with random prompts, I hate needing to use Discord for it... Is that 250 Prompts, since each prompt comes with 4 images, or 1000 unique prompts?
|
# ? Feb 21, 2024 14:39 |
|
pixaal posted:Why can't I just use it from the get go if I'm paying? Sab669 posted:Is that 250 Prompts, since each prompt comes with 4 images, or 1000 unique prompts? I've had alpha access for a while, and honestly, I think I prefer Discord generation over website generation still. The website seems slower to generate on average, and the ability to just recall my last prompt in discord with the up arrow on my keyboard / quickly paste an image to use for a prompt are a big enough benefit that I've yet to bother changing over. For anyone curious about the UI:
|
# ? Feb 21, 2024 15:06 |
|
The aspect ratio, weirdness and variety(chaos?) sliders are cool, I'm using it most of the time now. the UI could be improved. it has image handling for reference images but is kinda wonky about it
|
# ? Feb 21, 2024 15:10 |
|
Is parity 1:1 yet with discord? I'd rather not use Discord but I've been poisoned by muscle memory and memorizing all the little codes. Last I checked there were things the website couldn't do that discord could.
|
# ? Feb 21, 2024 15:15 |
|
There's an AI picture of Elon sleeping on SpaceX floor going around and I wanted to have his shadow demon drag him into the shadow realm, Pika decided this was stupid and made up it's own thing, typical pika. I open I get more control in OpenAI's video. I guess it works okay as a shapeshifter animation? https://i.imgur.com/wrl831j.mp4
|
# ? Feb 21, 2024 16:51 |
|
Yeah, I'm going to second (or third?) that discord is actually more powerful and less annoying than the site, which is actually a telling statement about how bad the UI is.
|
# ? Feb 21, 2024 17:01 |
|
A random grab-bag of stuff I've made in
|
# ? Feb 22, 2024 09:16 |
|
https://x.com/StabilityAI/status/1760656767237656820
|
# ? Feb 22, 2024 15:02 |
|
Ffs, I'm still not done loving around with stable cascade and now this. Slow down a little. Take it easy
|
# ? Feb 22, 2024 15:50 |
|
Maybe VRAM is lower than cascade. There can be different tools for different hardware! It's kind of where this is probably going.
|
# ? Feb 22, 2024 16:01 |
|
drat how are any of you getting text on stuff, are you all cheating and adding it in post?
|
# ? Feb 22, 2024 16:01 |
|
Elendil004 posted:drat how are any of you getting text on stuff, are you all cheating and adding it in post? On Dalle 3 talk to it with sentences and at the end. overlaying the image with gorgeous text reading "I am a goon" On Stable Diffusion you need that 3.0 that they just announced and haven't made public I think MJ does text pretty well too. SDXL does a good job at single words do not go for a sentence there. e: you can also do it more basic like this https://www.bing.com/images/create A birthday party with a banner saying "OH NO WARIO!". Wario is sad and smelly alone in the dim light e2: Google AI Kitchen made this if you like bashing your head against a filter that bans words like sitting, clothes, wearing, garb, and so on. I find it more a challenge to get it to run than useful. last time I posted how I made something in test kitchen half the words were blocked within 24 hours so I have to assume google is watching. pixaal fucked around with this message at 16:08 on Feb 22, 2024 |
# ? Feb 22, 2024 16:03 |
|
Elendil004 posted:drat how are any of you getting text on stuff, are you all cheating and adding it in post? With both Midjourney and SDXL I generally just make the main thing then in paint the text on that. It typically doesn't take too long to get what you want. V6 is much better in Midjourney than v5, SDXL is better than both. Short things like mentioned, titles, names, signs with a few words max work best for obvious reasons. If you have Stable Diffusion running within Photoshop then you literally just slap down your text and 'mesh' it with the main image using smaller amounts of steps. Same concept can work with various Stable Diffusion front ends, just they aren't Photoshop so you'd have to draw by hand which isn't too bad these days. For Midjourney you can't actually draw yourself so instead it's just inpaint roulette. EDIT https://t.co/lX8z1UTc8s Here is the link directly to Stability AI's Stable Diffusion 3 announcement KakerMix fucked around with this message at 18:03 on Feb 22, 2024 |
# ? Feb 22, 2024 16:11 |
|
Love this
|
# ? Feb 22, 2024 16:49 |
|
The taco bell! Open AI give me access to your video model please. https://i.imgur.com/Qis91jy.mp4
|
# ? Feb 22, 2024 22:14 |
|
|
# ? Feb 22, 2024 23:26 |
|
https://fastsdxl.ai Impressive speed.
|
# ? Feb 23, 2024 07:51 |
|
BoldFace posted:https://fastsdxl.ai The good: incredibly fast. Much better image quality than I can get running sdxl local. The bad: it's still sdxl. Prompting is completely different than midjourney, and I have yet to find an MJ prompt that transfers over cleanly. Descriptions of characters, poses, and costuming often get jumbled up with each other. Plus the ux is clearly just slapped together. Still, it's incredibly impressive. I might try using it to create image references for MJ, it's so drat fast.
|
# ? Feb 23, 2024 12:38 |
|
You cannot expect to port prompts 1:1 between AI they are trained differently and react to input different. Learn to talk to all of them. Why would you expect other AI to switch from the prompting style their users are used to and scrap their entire training workflow to get it to prompt like mj?
|
# ? Feb 23, 2024 13:07 |
|
Your post reads like it was created by Chat GPT. The circle is complete.
|
# ? Feb 23, 2024 13:13 |
|
you've been staring into the abyss too long https://i.imgur.com/flzheb6.mp4 pixaal fucked around with this message at 18:22 on Feb 23, 2024 |
# ? Feb 23, 2024 14:45 |
|
null_pointer posted:The good: incredibly fast. Much better image quality than I can get running sdxl local. if you want a good ux for stable diffusion stuff, i recommend diffusion bee. you'll still have to learn sd style prompting of course, but its a good app, easy to use, and free edit: sorry forgot that its ios only but i think there are some good ones for other platforms too
|
# ? Feb 23, 2024 18:48 |
|
Earwicker posted:if you want a good ux for stable diffusion stuff, i recommend diffusion bee. I don't think anything else supports lighting yet that's a preview barebones demo. that said I seem to remember similar for turbo so only a matter of time.
|
# ? Feb 23, 2024 18:55 |
|
null_pointer posted:Your post reads like it was created by Chat GPT. The circle is complete. The longer answer is the way you prompt in MidJourney will never be like you prompt in Stable Diffusion, XL, 3, or any version that comes afterwards because they are fundamentally different models. Likewise DallE will always be different as well, since it too is a different model. There might be vague similarities but really, each model (even within lines, v5 MidJourney is quite different than v6 than v4, same with Stable Diffusion 1.5 vs. XL). It isn't anything you are doing wrong, really, it's just that's how it is. Prompts are never 1:1:1 between any model. Stable Diffusion models on account of their open source nature will be more similar between each other because they all have the same base. Even then, anime models don't prompt like non-anime models and that's even within the same versions. For those Auto1111 lovers out there: https://github.com/lllyasviel/stable-diffusion-webui-forge Literally 'not poo poo' Automatic1111 with a whole heap of fixes, in-built extensions and quite a bit more efficient for lower end GPUs. I've completely replaced standard Auto1111 with this Forge version. It's identical, same extensions and everything you are used to, it's a 1:1 replacement, no downsides that I have found.
|
# ? Feb 23, 2024 19:24 |
|
|
# ? Feb 26, 2024 21:07 |
|
Every single one of those could be a new Dream Theater album cover
|
# ? Feb 26, 2024 21:15 |
|
|
# ? May 28, 2024 07:58 |
|
haven't listened to these guys before, but now im rockin images and words
|
# ? Feb 26, 2024 22:04 |