- Joined
- Aug 17, 2018
Stable Diffusion, NovelAI, Machine Learning Art - AI art generation discussion and image dump
- Thread starter Near
- Start date
- Joined
- Apr 14, 2021
I noticed interesting with DALL-E 3 when using specific names. In this case, I gave it a prompt describing this image to see how close a result I could get:
For some reason he vaguely reminded me of an actor, so I decided to replace "a man" in the prompt with the actor's name. Despite generating faces that looked nothing like him, it seems to have figured out his ethnicity based on his name with no problem:
For the next one, to try to make it give the right face, I edited the prompt to contain "the famous actor [name] who starred in x movie". The resulting image still looked nothing like him, but it seems to have realized the genre of movie being referenced, because it went out of its way to have a part of his shirt entirely missing to display gang tattoos:
While this was a yakuza movie, the actor in question didn't actually play a yakuza character, so it seems like it did take the mentioned movie into account while also completely ignoring the person's name.
I'm legitimately fascinated by this result and I don't know why it does this. Is DALL-E 3 made to ignore people's names specifically? Also, it seems like this movie doesn't have an English synopsis on IMDB nor an English page on Wikipedia, so is it drawing from foreign language sources despite not accepting non-English input?
I'm legitimately fascinated by this result and I don't know why it does this. Is DALL-E 3 made to ignore people's names specifically? Also, it seems like this movie doesn't have an English synopsis on IMDB nor an English page on Wikipedia, so is it drawing from foreign language sources despite not accepting non-English input?
- Joined
- Aug 17, 2018
- Joined
- Aug 22, 2015
I always thought he had a little resemblance to Matthew Perry so I tried generating the image with that and got instantly blocked. Thanks Bing, very cool. Prompt works for just generic "man", but Perry's name got me a no-no warning.
- Joined
- Aug 2, 2019
THIS SUMMER
>BWOOOOM
>"THEY TOOK MY FUCKING SISTER, VANESSA"
ONE MAN
>dramatic zoom on protag
WILL SPEND A WORK WEEK
>scene of protag chambering a round
>"It's on!"
CLEANING UP THE SHOP
>BWOOOOM
>"THEY TOOK MY FUCKING SISTER, VANESSA"
ONE MAN
>dramatic zoom on protag
WILL SPEND A WORK WEEK
>scene of protag chambering a round
>"It's on!"
CLEANING UP THE SHOP
I swear to God I was going to throw in a Living Tombstone joke but the madmen actually threw that in credits
Last edited:
- Joined
- Aug 17, 2018
Why can't we have statues like her?
- Joined
- Jan 31, 2021
Forgive my bluntness, but it's because she's not a nigger.
- Joined
- Oct 20, 2018
- Joined
- Dec 17, 2019
- Joined
- Apr 14, 2021
Please be patient, the moron who owns this account is experiencing technical difficulties (read: boomer with technology)
Last edited:
- Joined
- Apr 14, 2021
This has become the most absurd hobby, but damn it's fun
I told it to put baby faces on the cookies, but the Charlie Chaplin look outfit made it too funny to not keep anyway:
In honor of our Erverlord, pitty's pizza day:
And lastly, one that I'm going to need to touch up to produce something funnier (tried to have a pile of empty baby shoes next to the pitbull):
I'm really curious why it blocks Matthew Perry but Show Aikawa was fine, did Perry not have any Good Goy Points?
I told it to put baby faces on the cookies, but the Charlie Chaplin look outfit made it too funny to not keep anyway:
In honor of our Erverlord, pitty's pizza day:
And lastly, one that I'm going to need to touch up to produce something funnier (tried to have a pile of empty baby shoes next to the pitbull):
That really is a great image, he has the same vibe as the original lol
I'm really curious why it blocks Matthew Perry but Show Aikawa was fine, did Perry not have any Good Goy Points?
Last edited:
- Joined
- Feb 5, 2023
I'll recommend some text2img A.Is that I liked. I'm dumb and I've made an entire thread for it some minutes ago because I didn't noticed this one.
Pixai.art
This is my favorite so far. It's image generation is almost perfect in my opinion and it follows the prompt very well. It also has an option to animate the image, but currently the animations are pretty crappy. The only problem is that you need a subscription after generating 3-4 images, but you can bypass this by just creating a new account, and tempmails works on that occasion, use Yopmail.
Bing AI
A really controversial one. It's generation is also very good, but... Microsoft. Just use it if you're too lazy for creating many sockpuppets on Pixai.art or if you don't want an anime pic.
Stable Diffusion 2.1
This is a demo of Stable Diffusion. It's really weak compared to the original one, but it works if you need something simple. I highly recommend hosting a local server with the original Stable Diffusion, but if your GPU can't handle it you can use this demo.
Frosting.ai
I liked this one. You don't really need an account for generating and it's generation is fast. Although there are some shitty restrictions, for example if you want to generate an image of an old guy, you can't because it isn't allowed to specify an age on the prompt.
There are many others which I haven't tested yet:
https://www.mage.space/
https://tensor.art/
https://playgroundai.com/
https://getimg.ai/
https://www.patience.ai/
https://beta.dreamstudio.ai/
https://dezgo.com/
Pixai.art
This is my favorite so far. It's image generation is almost perfect in my opinion and it follows the prompt very well. It also has an option to animate the image, but currently the animations are pretty crappy. The only problem is that you need a subscription after generating 3-4 images, but you can bypass this by just creating a new account, and tempmails works on that occasion, use Yopmail.
Bing AI
A really controversial one. It's generation is also very good, but... Microsoft. Just use it if you're too lazy for creating many sockpuppets on Pixai.art or if you don't want an anime pic.
Stable Diffusion 2.1
This is a demo of Stable Diffusion. It's really weak compared to the original one, but it works if you need something simple. I highly recommend hosting a local server with the original Stable Diffusion, but if your GPU can't handle it you can use this demo.
Frosting.ai
I liked this one. You don't really need an account for generating and it's generation is fast. Although there are some shitty restrictions, for example if you want to generate an image of an old guy, you can't because it isn't allowed to specify an age on the prompt.
There are many others which I haven't tested yet:
https://www.mage.space/
https://tensor.art/
https://playgroundai.com/
https://getimg.ai/
https://www.patience.ai/
https://beta.dreamstudio.ai/
https://dezgo.com/
- Joined
- Aug 22, 2015
There might be some automated thing that blocks recent newsworthy stuff, I don't know.
- Joined
- Apr 14, 2021
I just found someone earlier in the thread say that they block anything they can find a recent news article for, I assume so they can make sure no "controversial" memes can come out of it
- Joined
- Dec 27, 2019
Played a few rounds of Call of Duty and got myself thinking what it would look like with kiwis.
These are kind of cute.
These are kind of cute.
ExsanguinateHorizon
kiwifarms.net
- Joined
- Dec 16, 2019
I love the chaotic doodad clouds. The little fellows even have trigger discipline.
- Joined
- Jan 31, 2021
Artflow finally reveals the new skinny legend, Null.
I tagged him initially, but I don't want to get featured again on MatI but this time on how to get instabanned.
And a not so skinny legend
I tagged him initially, but I don't want to get featured again on MatI but this time on how to get instabanned.
And a not so skinny legend
lilmayo420
kiwifarms.net
- Joined
- Jul 17, 2022
Lord of the Pajeets
kiwifarms.net
- Joined
- Mar 27, 2018
Love me rats. Civit scrapes and displays all the metadata from the images in their gallery, so if you view the images for whatever version of the LoRA it will give you a good path towards making your own (e.g. https://civitai.com/images/3258873). I mainly use Dreamshaper 8 for the SD1.5 LoRA and Dreamshaper XL for the SDXL version (with varying degrees of https://civitai.com/models/127139?modelVersionId=163652 mixed in).
