D&D General DALL·E 3 does amazing D&D art

[Parmandur]

Asked Bing Chat to describe the Caves of Chaos, and then draw a band of Adventurers going in:

 

[Zaukrie]

Asked Bing Chat to describe the Caves of Chaos, and then draw a band of Adventurers going in:

View attachment 318323View attachment 318324View attachment 318325View attachment 318326

When I used band in prompts, I got musicians every time....

 

[Parmandur]

When I used band in prompts, I got musicians every time....

Exact prompt:

A band of adventurers entering the canyon of the Caves of Chaos

Maybe you have to specify "band of [x]"?

 

[Zaukrie]

Exact prompt:

A band of adventurers entering the canyon of the Caves of Chaos

Maybe you have to specify "band of [x]"?

It never worked for me.... But I'm glad it did for you

 

[Parmandur]

Prompt = style is romantacism; norse god odin sitting on a throne; odin is holding his gungnir spear; two wolves lay at Odin's feet; two ravens perch on Odin's shoulders

Spoiler: View Images

1
View attachment 318204
2
View attachment 318205
3
View attachment 318206
4
View attachment 318207

1950's Hollywood Technicolor Epic



Telegu film



Wuxia

 

[Saracenus]

Except that kinda defeats the point: if I already had an image of a silver dragon the odds are high I wouldn't need another one.

It's things for which there are no example images (e.g. two teams of mostly Dwarves playing hockey, with the teams' uniforms being of specific colours) where this AI art could be fun and-or useful, and I'd like to think it's smart enough to interpret the prompts I type in rather than having to rely on AI to generate those prompts for me.

From what y'all have been saying and displaying, it seems it still has a way to go on interpreting prompts; which is fair enough as this is all still in its relative infancy.

We are light years ahead of where we were when the first DALL-E came out and even DALL-E 2 is almost like cave drawings from what we are doing now. We are already seeing improvements in the data because unless you turn of the data sharing in GPT-4 your prompts and evaluations of the output (that is what the thumbs up and down icons are for) it is learning as more people work with it. The most obvious change in the past few weeks is fingers, it doing a much, much better job with fingers... Eyes are still a work in progress.

BTW, I just saw a post from someone who is using GPT+ like I am and they are rolling out a more integrated version of GPT-4 slowly to us paid users (I have not seen this version yet) where it can read uploaded PDFs and analyze them for you, it will contextually use Search with Bing or DALL-E 3 or Advanced Data Analysis based on what you are asking it to do. She showed an example of uploading a photo and creating an image based upon it and then uploading a different image to add to it... LINK: @luokai on Threads
EXAMPLES from her post on Threads:



I think this is going to make GPT+ a killer app that is worth the price of admission. Right now I can enable Search With Bing or DALL-E 3 or Advanced Data Analysis, but not all at the same time. Hopefully I will get this new, improved version in the coming week or two and I can report back on how it is working.

I will say, I am happy with the capability I have now for images and I am just beginning to explore the other features. One I am very interested in trying out is the integrated VISIONS in default GPT-4 because I am barely tapping its capabilities, for instance I can take a photo of my refrigerator contents with my phone and have visions catalog it and suggest recipes to use what is in there.

I am not sure which beta plugin does it yet, but you can have GPT evaluate your website or WordPress blog and it will help you streamline the design, make it more visible to search with metadata, and create all kinds of improvements that would require a lot of learning or a lot of money to a web dev.

 

[Parmandur]

Something is stirring

 

[Parmandur]

Jane Austen Dungeonpunk party

 

[pukunui]

I got my 15 boosts, and I've been using them to try and make an image of my daughter's kalashtar PC, Moonflower, with her quori spirit. This is the closest I've gotten so far:

Prompt: Detailed ink and watercolor. Forest glade setting. Full length view of female human sorceress with shoulder-length silver hair, dark tanned skin, and purple eyes. She is wearing a medieval blouse and skirt with leather boots. She smiles as she strides forward, her boots crunch on fallen leaves and sticks. Behind her looms a translucent blue spirit monster with lobster claws.

 

[Saracenus]

Enstad serves as the capital of the Kingdom of Celene, located to the west of the City of Greyhawk and southwest of the City of Dyvers. Situated at the headwaters of the Handmaiden River, the town employs both its strategic natural geography and potent elven magic for defense. The elves of the Fairy Kingdom of Celene disdain stone walls, which they believe severs their connection to the life-giving energies of the natural world. At the heart of the capital stands the Queen's palace, an architectural wonder crafted from living trees that have been magically shaped, epitomizing the essence of the town.

GENERATED PROMPT: Detailed Illustration: A breathtaking aerial photograph of a fantasy elven town, Enstad, during the golden hour. The city's organic layout harmoniously integrates with the lush landscapes and the serene Handmaiden River flowing from North to South. The sparse buildings appear as if they have naturally risen from the earth, with roofs covered in moss and walls made of intertwined roots. The Eastern bank of the Handmaiden is dominated by forest with a few dwellings at its edge. In the center of the city is the palace of the Queen, made of living redwood trees shaped by magic, and stands as a mesmerizing focal point. The city subtly radiates a magical glow, with the river reflecting the city's beauty. The predominant colors are golden yellows, deep greens, and shimmering blues.

784 characters, welp I still haven’t found the limit to GPT prompts. Over double the limit of Bing.