#dnd style portrait

1 messages · Page 1 of 1 (latest)

mellow bear
#

First test with instructions given by @mild gale
Ok not to beat the dead horse.. but I found my exact prompt :

Can you creat this character?

Marta is a teifling. She has distinct purple skin that catches the eye, complemented by her long silver hair that cascades down to her waist. Her striking golden eyes exude curiosity and intelligence. She has small, well-maintained horns that curl around her face, and her tail sways playfully behind her. Marta dresses in a mix of practical and stylish clothing, often adorned with accessories that showcase her eclectic tastes. She is an achedemic at heart but runs a general store in a swamp.

mild gale
#

wait why does yours look so much better than mine! You got the store and everything lol

mellow bear
#

I tried fresh chat without CI

tacit elk
#

@mild gale Since you're limited to bing chat, I wonder if you've experimented with this tool, and if you haven't, if it would work for you better than relying on Bing Chat's ability to make the correct prompts for DALL-E:

https://www.bing.com/create

mellow bear
#

I did mine with chatgpt

mild gale
#

I think its very nice picture

tacit elk
#

For the bing image creator, you feed it direct DALL-E prompts, rather than bing chat making the DALL-E prompts for you based off your overall request.

#

Give it a shot with that summarized/optimized prompt I gave you near the start of our convo

#

(The version of it that includes a brief description of the art style we were shooting for)

mild gale
#

Guys I did it

#

the ultimate protrait

tacit elk
#

Using bing chat? Or the image creating tool

#

Oh no, the use of the word ultimate makes me fear impending sarcasm lolol

mild gale
tacit elk
#

KNEW IT

mellow bear
#

What is that

mild gale
#

Its beautiful (Sorry we are derailing your stuff neko)

tacit elk
mellow bear
#

🥲

tacit elk
#

Oooooh. It's not exactly what we were looking for, but I just pulled this off with ChatGPT's DALL-E module

#

Not head-to-toe, but thats the closest i've gotten progress-wise so far getting DALL-E to replicate the desired art style

mellow bear
#

I should try vision with @mild gale input

mild gale
#

Man Bing is acting weird today

#

I asked it to generate an image and it extended out the text instead

#

Then created an image:

tacit elk
mild gale
#

The other website?

#

As soon as I loaded it it said "Our engineers are looking into solving the problem"

tacit elk
#

Its the same website (bing), just a tool that lets you feed prompts directly to DALL-E rather than using bing chat to try to guess how to best make the prompt for you

mild gale
tacit elk
#

What in the world

#

I'll give it a shot, assuming they still temporarily give everyone free credits

mild gale
#

The create button is greyed out

rugged cipher
mild gale
#

Its weird because bing Chat makes suggestions like "Can you make the background Martha's shop?" but if you click it, it just ignores the prompt lol

#

I dont need the feature, I just think its funny it suggests them

tacit elk
# rugged cipher

Whoooooooooaaa... I'm still a bigger fan of the more cell-shaded hand-drawn pencil-ish style we were going for earlier, but still.... whoooooa... easily 2nd place from all the generations I've seen for it today

tacit elk
mild gale
#

Everytime I try to make the prompt better, it gets worse 😦

tacit elk
#

So, I don't know if I clarified it earlier

#

but when I first was telling you about summarization/optimization for DALL-E, that was wrongly thinking you were using the bing image generator to begin with

mild gale
#

Oh

#

i've been trying to shortened version

tacit elk
#

If you're using ChatGPT or Bing chat, your original method is the better method, because bing chat uses its version of GPT to turn that request into what it perceives to be the best prompt for DALL-E based on your request

#

While the bing image generator makes no optimizations to the inserted prompt, it just uses DALL-E to generate with the exact direct text you feed it

mild gale
#

Yeah but your right, it ignores a lot of important details. Like "Black hair in a bun" turned into White hair sholder length with a guatiar?

tacit elk
#

Yeah dude, look at this

mild gale
#

This one is kinda nice

tacit elk
#

First try with image creator tool

mild gale
tacit elk
#

Im sure with repeated use it would create one you'd wanna keep as the main one

mild gale
#

Yeah its still throwing the error for me

tacit elk
#

But it was my exact prompt I gave you earlier that succeeded first try with the "full-body" part, cause it wasn't GPT guessing what the prompt should be, it was me making sure the most important parts were there by feeding it directly myself

#

Are you in USA and not using a VPN? (I have no basis as to why that would be a reason of it not working, other than the fact that those two things are consistently present for users who never run into any issues with anything OpenAI related)

mild gale
#

I am in US and not using vpn

tacit elk
#

I retract my statement then, lol

mild gale
#

Honestly, I need to get back to actual prep soon 🙂

#

I did have a fun side track on this

tacit elk
#

I'll use the image generator tool with that prompt to generate a few more sets of images and I'll send them all here for you, maybe one will work well for you while you focus on the stuff you need to focus on

mild gale
#

I think maybe the system is just overloaded because now its just throwing my prompts away completely

#

Prompt I submitted:

#

Prompt it used

tacit elk
#

"She thinks shes better than you" eh, she probably is

mild gale
tacit elk
#

Love that you described her desired expression that way, though lol

mild gale
#

Yeah but it discarded everything but her name (which it never has done before)

tacit elk
#

You should be clarifying in that first question more specifically of the kind of portrait you're requesting

#

I know it was just the mention of a D&D character that was working with you before, but for shorter summarized requests like that, without getting real specific on the style, its going to levitate toward photorealism/actual realism 90% of the time (In my experience, especially with humans and any non-fiction creature/animal )

rugged cipher
mild gale
rugged cipher
#

And I would like to remaind that Dalle is not Midjourney, here you can speak with Model and don't have to give exacly precise prompts but develop together.

analog lotus