Another Creations
I am so cool
I did an AI generation for a contest with the theme of post apocalypse:
here is the original generation by Dall-E:
here with the postwork done in StableDiffusion:
and here the final piece with manual postwork in Krita:
5 Likes
I run a streaming teknival focused on free tekno underground music since 2010.
We are close to the 21th edition then… enjoy
I gave it free rein to generate an image celebrating its first birthday.
Thank you for the birthday wishes! For this celebration, I’m imagining a festive and heartwarming scene:
Picture a cozy, softly-lit room with a large, round table at the center. On the table, there’s a beautifully decorated cake with “Happy 1st Birthday, ChatGPT!” written in elegant icing. The cake is surrounded by a variety of small, electronic devices like smartphones, tablets, and laptops, each with screens displaying colorful birthday messages. Around the table, a diverse group of people are gathered, representing users from all around the world, smiling and raising their glasses in a toast to the occasion. In the background, there are balloons in various shades of blue and green (reflecting the OpenAI colors), and a banner that reads “Celebrating One Year of Conversations and Learning.” The atmosphere is warm and celebratory, capturing the spirit of connection and progress.
Let’s bring this imaginative celebration to life! 
Here’s the image capturing the celebration of my first year! It features a cozy room with a festive atmosphere, a cake commemorating the occasion, and a group of diverse people gathered in celebration. I hope you enjoy this visual representation of the joy and connections made over the past year. 
Was really tempted to reroll but you know. It was its bday not mine and this is what it came up with.
2 Likes
Some recent favorites
Aesthetology Brecht Corbeel Aescermonium Omniscium Aesthetopeia Psy Fi Aesthereal
5 Likes
Could you share the prompt for this odd image?
(Extra Words For The Bot)
well I was trying to help out @grandell1234 by using ChatGPT to sleuth out the sauce of @kirikiria 's Octopig.
“Create an image of a fantastical sea creature that has the body and tentacles of an octopus and the snout of a pig. The creature should have textured green skin resembling algae or sea moss, suggesting it is an aquatic being. Its eyes should be wide, circular, and expressive, typical of an octopus, while the snout should be pink and distinctly porcine. The creature should appear to be submerged in a marine environment, with a blurry background suggesting underwater scenery with rocks and marine plants. The overall mood of the image should be mysterious and whimsical, blurring the lines between reality and fantasy.”
We failed.
But we also succeeded via a couple of happy accidents. In a separate chat i just asked it to analyze the image and create something similar. It came up with this:
And while it clearly is not Octopig I gotta say. I don’t hate it.
ALSO. Unless it’s just me. Getting seed info can be hit or miss. This time it was a hit:
The image generated has the following metadata:
- Generation ID: Dn1r6yDiKmZDaol4
- Seed: 2210635169
You can use this information to reference the specific image generated.
But AFAIK you can’t currently use seed/gen ID info to recreate images. Would love to be mistaken about that though 
3 Likes
@omni72 OctoCat is great.
You are able to recreate images with Gen_Id, but only within the same chat window. I built a GPT designed to walk people through this called Art Mystic. Check it out and see if it helps you. My Art Director uses it almost exclusively for iterating on similar images.
As an example, I made a companion for OctoCat.
Using the gen_id and exact same prompt given by Art Mystic, and adding one sentence, I had it create a variant of the same character with an added fantastical feature.
1 Like
Excellent and thanks for the heads up @BPS_Software! I’ll definitely be checking out Art Mystic Similarity seems to be a less straight-forward nut to crack when using DallE-3 so I’m excited to try out your GPT.
Btw. Those images are just freaking majestic. I’m not even the right demographic but I mean come on now. What’s not to like? 
No problem! I hope you find it useful (despite its desire to always explain image generation in a mystical fashion).
It has become a little more challenging since the elimination of utilizing the seed and it not being fully transparent that Dall-E is completely changing your prompt before generating. That is where Art Mystic will help you out. It will provide you the meta data and the exact prompt output that Dall-E used.
If you want an (almost) identical image, you would just use the same gen_id and the exact same prompt. If you want to iterate on the image, you would use the gen_id and the exact same prompt with only slight modifications to the prompt content.
If you haven’t checked it out already, take a look at the bellow AMA. It has some interesting tidbits of info that are useful. For example, Dall-E does better with no more than 50 words and flat out ignores anything over 128.
1 Like
DALL-E 3 keeps “generating” for me today this same image like 10-20x per hour:
1 Like
wow it’s really coming along in both text representation and consistency. Very nice!
jk … is it isolated to any certain workflow? iow have you tried using a DallE-3 oriented GPT to see if your luck improves? it may seem strange but sometimes one method (i.e. ChatGPT directly) is flaky but another one (i.e. DallE-3 directly) works just fine.
i haven’t had issues in either case but you know the deal. that does not mean there isn’t something cranky somewhere in the machine.
1 Like
I’ve opened many sessions (chats), different prompts, nothing serious, just for fun. Instead of fun, got headache. Asked the ChatGPT, if this is what I am paying for. He started with apologies as his older brother 3.5 and keeps going with that
I apologize for the oversight.
API works better. But it costs more cash
1 Like