|
That is her resting-think face. Don't shame
|
# ? Aug 5, 2022 18:45 |
|
|
# ? May 30, 2024 03:15 |
|
A few results from my current DALL-E 'project' (Topiary of two hands pulling something apart from the outside, with a ring on the left hand) (Two hands stretching open a knot in a tree)
|
# ? Aug 5, 2022 19:22 |
|
Stable Diffusion email invites have been sent just now
|
# ? Aug 5, 2022 19:45 |
|
So far i think midjourney makes prettier stuff when it comes to digital art, but otherwise lacks in customization and ability to differentiate between two subjects matter compared to dalle2 Dalle2 is basically a bit trickier to use well. But, as a game dev, my team is for sure looking at mid journey for our needs since gently caress the payByGeneration system of Dalle2. Can’t beat 30/50 a month for unlimited usage.
|
# ? Aug 5, 2022 19:54 |
|
KinkyJohn posted:Stable Diffusion email invites have been sent just now darn, missed the boat on it
|
# ? Aug 5, 2022 20:04 |
|
https://twitter.com/imperishablneet/status/1555660485118271491
|
# ? Aug 5, 2022 22:36 |
|
KinkyJohn posted:Stable Diffusion email invites have been sent just now Thanks for this, the email was in my spam folder. Hanging out in the discord and there are a few things clear. One, in 20~ hours the discord bot goes live. 2, absolutely no NSFW images allowed BUT they are also very quick to also say 'when the local version is released you can do whatever you want, no NSFW on discord though'. 3, celebrities are absolutely allowed and there seems to be zero restriction on anything otherwise, especially compared to Dall-E. The images the staff are sharing certainly look really great too, so much so that I don't know if Dall-E can compete.
|
# ? Aug 5, 2022 23:03 |
|
Darn I didn’t even realize I could have signed up for that one
|
# ? Aug 5, 2022 23:09 |
|
StableDiffusion invite was in my spam folder too, time for some more bread stick rats and seals (after the 24 hour waiting period).
|
# ? Aug 5, 2022 23:10 |
|
Snowy posted:Darn I didn’t even realize I could have signed up for that one
|
# ? Aug 5, 2022 23:24 |
|
Son of Rodney posted:A dark field during a storm with lightning striking with an ominous, gigantic pig in the background I used this prompt in Midjourney Other Midjourney ones: 'two dragons fighting, by picasso' 'the most beautiful image ever created' 'the world's angriest hotdog with human eyes' 'thursday blues on the moon' 'Three Rings for Elven-kings under the sky'
|
# ? Aug 6, 2022 00:16 |
|
No specific images noted as such, but as a sign of the times, here's a kickstarter for an RPG campaign book that plans to use Midjourney as a tool for illustrations: https://www.kickstarter.com/projects/silveraxe/the-scourge-of-northland More Midjourney used to make comics, Ryan Estrada uses the classic A Softer World format to use the prompt as text for the output https://medium.com/@ryanestrada/for-exposure-comics-33ec400737ca fez_machine fucked around with this message at 01:37 on Aug 6, 2022 |
# ? Aug 6, 2022 00:55 |
|
I've never used discord before so I want to make sure I understand what's going on with stable diffusion; slow mode is enabled which means I can only post one prompt per hour, correct?
|
# ? Aug 6, 2022 01:36 |
|
In 24 hours rhe bot will come online afaiu (turns out my invite was in the spam folder)
|
# ? Aug 6, 2022 01:37 |
|
Tunicate posted:In 24 hours rhe bot will come online afaiu (turns out my invite was in the spam folder) Mine was too and i really appreciate everyone posting about it. I was able to post one prompt, and I see other people posting prompts, so I guess everyone gets 1 until the bot is online maybe?
|
# ? Aug 6, 2022 01:39 |
|
I got a prompt through on stable diffusion "Nicolas Cage in full costume and makeup as Darth Icky, new Star Wars villain, photo from Entertainment Weekly cover shoot" cage is looking rough lol
|
# ? Aug 6, 2022 01:56 |
|
Cage kinda looks like Bashar Al-Assad in that one…
|
# ? Aug 6, 2022 01:59 |
|
TIP posted:I got a prompt through on stable diffusion Is... is that a bionicle?
|
# ? Aug 6, 2022 02:10 |
|
Mercury_Storm posted:StableDiffusion invite was in my spam folder too, time for some more bread stick rats and seals (after the 24 hour waiting period). hell motherfucking yeah
|
# ? Aug 6, 2022 03:32 |
|
I'm continuing my effort to confuse AIs by seeing how tight they couple "red panda" as a concept with the color, Stable Diffusion is the best so far but I don't have Dall-E 2 access yet (prompt was "Ailurus fulgens with blue fur", using the latin name to make it a bit easier for the AI)
|
# ? Aug 6, 2022 03:51 |
|
To be real clear about Stable Diffusion: it isn't a 24 hour wait for people specifically, when those emails were sent out it was 24 hours from when the discord bot goes live for everyone. The discord is saying 2pm EST, as far as I know, tomorrow. EDIT They do have a clunky channel to ask for prompts that moves slowly, so you might be able to get one through that way. I'm just going to wait. In hanging out in the discord though it feels very much like a Dalle mini/craiyon but way more powerful and very clearly the best at faces out of all the Ai generators so far. KakerMix fucked around with this message at 04:04 on Aug 6, 2022 |
# ? Aug 6, 2022 04:01 |
|
KakerMix posted:To be real clear about Stable Diffusion: it isn't a 24 hour wait for people specifically, when those emails were sent out it was 24 hours from when the discord bot goes live for everyone. The discord is saying 2pm EST, as far as I know, tomorrow. Wait, is there a way to do Stable Diffusion prompts without going through the request channel?
|
# ? Aug 6, 2022 04:12 |
|
Got into DALLE2 and have been going a little crazy with it. From the ProcGen thread:Lord Hydronium posted:Got access to DALLE after about 7 weeks, and created this monstrosity: “A velociraptor playing electric guitar at a concert, vaporwave” “A stained glass window depicting the woman yelling at cat with salad meme” “Impressionist oil painting of the Burger King sitting on a hamburger shaped throne” “Marble sculpture of Fred Flintstone on a dark background” “Renaissance oil painting of the Hamburglar at the birth of Christ” “Renaissance oil painting of a grey alien and angels” (this was somehow the least creepy of the images it generated) “Cave painting of three people waving spears at a UFO” “Sumi-e painting of Godzilla attacking Tokyo”
|
# ? Aug 6, 2022 04:17 |
|
ymgve posted:Wait, is there a way to do Stable Diffusion prompts without going through the request channel? Some people have access but they are special. I am not special. Tomorrow it will be like MidJourney in that its a bot you can just ask to generate things for you.
|
# ? Aug 6, 2022 04:26 |
|
Rutibex posted:i wonder how long it will be before they teach an AI to program games and this can be real. From several pages back but I think the intermediary step is going to be stuff like this which is also really exciting: https://www.youtube.com/watch?v=udPY5rQVoW0
|
# ? Aug 6, 2022 05:13 |
|
In trying to find that video I also found this which is pretty cool. Some guy made a self driving car in GTA that drives at least as well as I do in GTA, with some minor hiccups here and there, presumably based on image recognition. https://www.youtube.com/watch?v=rvnHikUJ9T0 Does anyone have any good videos of neural network algorithms playing videogames? Playing videogames is a type of art, right?
|
# ? Aug 6, 2022 05:22 |
|
|
# ? Aug 6, 2022 05:35 |
|
Getting faces that look a bit wonky? You can try running your output through these face fixers - ARC: https://arc.tencent.com/en/ai-demos/faceRestoration gpen: https://replicate.com/yangxy/gpen
|
# ? Aug 6, 2022 08:32 |
Ceramic Shot posted:I'm getting pretty good at prompts, but I've got a question for someone with more Midjourney experience: how do you prompt such that it's more likely to understand you want two distinct characters instead of an amalgam of your terms? No matter how I try to make them dreadfully distinct with interacting verbs, the image produced is often a melding. I believe you separate with :: as opposed to a comma. They discuss it in the midjourney docs Dr. Fraiser Chain fucked around with this message at 08:49 on Aug 6, 2022 |
|
# ? Aug 6, 2022 08:45 |
|
Goodpancakes posted:I believe you separate with :: as opposed to a comma. They discuss it in the midjourney docs Thanks! I found the relevant text in the FAQ, the third tab of the documentation, not in the User Manual if anyone was curious. Here it is: Q: Do commas, pipes, or any other punctuation matter? A: Yes, the model considers them like a break in the prompt, but it's not always clear what impact this will have. Use :: between prompts for a hard separator as describe in the answer above. And here's the example they mention: [image URL] [text prompt 1]::2 [text prompt 2]::3 --iw 1 results in a 1/6=17% image(s) , 2/6= 33% text1 , 3/6 = 50% text2. Though it seems like to get the "::" to work, you might have to be using an image URL? It doesn't seem to work with just text prompts, but I could be wrong. Edit: found a really nice Google Doc about how to better use inputs with Midjourney: https://docs.google.com/document/d/1XUT2G9LmkZataHFzmuOtRXnuWBfhvXDAo8DkS--8tec/edit Ceramic Shot fucked around with this message at 09:23 on Aug 6, 2022 |
# ? Aug 6, 2022 09:17 |
|
KakerMix posted:Thanks for this, the email was in my spam folder. Same! Almost missed this, the invite is only good for 24 hours. Check your spam for the discord link! edit: it says they sent it out to everyone that signed up. pixaal fucked around with this message at 10:56 on Aug 6, 2022 |
# ? Aug 6, 2022 09:51 |
|
Beartaco posted:From several pages back but I think the intermediary step is going to be stuff like this which is also really exciting: we have become dream thieves. harvesting the imagination of lesser beings like some hosed up fae
|
# ? Aug 6, 2022 10:26 |
|
Man I can't believe this prompt worked
|
# ? Aug 6, 2022 14:09 |
|
Beartaco posted:In trying to find that video I also found this which is pretty cool. Some guy made a self driving car in GTA that drives at least as well as I do in GTA, with some minor hiccups here and there, presumably based on image recognition. That is basically every video this person puts out: https://youtube.com/c/CodeBullet My favorite one is trying to teach a block character to walk to the right was slow. Then he increased the training rate by putting a death laser behind them. https://youtu.be/K-wIZuAA3EY Whichever block character survived the longest (usually by falling the correct way) made it to the next generation. EVIL Gibson fucked around with this message at 14:32 on Aug 6, 2022 |
# ? Aug 6, 2022 14:25 |
|
Wheany posted:Man I can't believe this prompt worked emojis are still text, which you can see if you hover over one, so you literally wrote "mushroom cloud"
|
# ? Aug 6, 2022 14:27 |
|
Beartaco posted:From several pages back but I think the intermediary step is going to be stuff like this which is also really exciting: I remember seeing this a year ago and being pretty impressed. After googling around though it doesn't seem like anyone is really taking up the reins on this, at least not publically. Other than this GAN Theft Auto video / GitHub project and Nvidia's original Pac Man experiments I can't find anyone who's working on simulating actual game environments with neural networks.
|
# ? Aug 6, 2022 15:27 |
|
This was just posted on the stable diffusion discord:quote:Last minute opening!
|
# ? Aug 6, 2022 18:12 |
|
Donatello does machines
|
# ? Aug 6, 2022 18:17 |
|
Zombie Killer 5000 posted:This was just posted on the stable diffusion discord: Thanks!
|
# ? Aug 6, 2022 18:19 |
|
|
# ? May 30, 2024 03:15 |
|
Zombie Killer 5000 posted:This was just posted on the stable diffusion discord: Thanks, I'm in! The actual system isn't live yet, right? Because I can't see a way to make more poo poo images so far. Disco Diffusion worked very well for generating more abstract, artistic images so I'm quite curious how this is gonna work out. I think at this point I have access to all the major systems, just Google holding back as always
|
# ? Aug 6, 2022 18:29 |