[MaxBigfoot]

Yeah, it looks like if you want to create a really good AI image, you still have to be a decent artist in the first place.   

[Nimrod]

Like others I have posted AI generated pics in this thread that I found interesting. I have also posted some comments from a few AI artists, about what it takes to produce an AI image that is what they had intended, whither real, or more fancifully, without some manual processing. Nimrod's comments here only prove that the state of AI, while a powerful tool, still has far to go until someone like me with little artistic skill can even get close to creating something that remotely looks real.        

Thanks for the acknowledgment Sam V. I do agree that to get the best out of AI art generation now, and possibly into the foreseeable future, one does need to have some artistic ability. Even if it is in the skilled selection of appropriate descriptors, or matching style to intent, the skill is still working with artistic tools.

[T_Man]

I just hope that someday they can combine this AI art with a 3D printer so they can create a real woman like this.

From yar1111 at DA.

[ghosthost]

Can I ask--what AI software is being used to produce fairly real looking mega-stacked babes? I've been experimenting with mid journey but they go berserk if there's even a tiny hint of sexuality in a prompt, like "big chest", etc. An exaggerated or naked breast would be impossible to create there, so...

[Nimrod]

Can I ask--what AI software is being used to produce fairly real looking mega-stacked babes? I've been experimenting with mid journey but they go berserk if there's even a tiny hint of sexuality in a prompt, like "big chest", etc. An exaggerated or naked breast would be impossible to create there, so...

There is no single perfect combination or software, although they all follow the same approach to generation.

Think of it as trying to tune in a radio station from a neighboring country. If the settings are off you will only get some Mexican radio and not the signal from Toronto.

It maters on:
The Software being used
The degree of control you have over the base algorithm (such as Stable Diffusion settings)
The limitations imposed by the software regarding NSFW content
The prompting seed TEXT or IMAGE (if any)
and most of all...
The library used to train the AI which not only imparts style but the ability for the AI to generate something. If it is not in the library the AI will struggle to "draw" the solution.

That is why you can often determine the library being used and not necessarily the software.
But that is only part of what it takes.

Even with the best of all the above running, if you can not provide a "conducive seed" then you are not going to get what you want/expect. Like trying to drive from New York to LA and starting by driving EAST.

My only suggestion it to work with the way you form your prompts, the manner in which you describe, and watch some examples. Most people run into problems because they are too direct with their seed text.

1. Avoid slang or jargon
2. Use more adjectives and fewer nouns
3. Describe the aspects and not the things
4. Use a deeper vocabulary while understand the subtle differences between synonyms

Realize that if the content is not in the teaching library the AI will be very hard pressed to create something because it really is not creating anything NEW, but rather the optimum solution that fits the prompts.

If you want total control, just download Stable Diffusion (free), find an interface app that you like, get a workable teaching library, and build up your vocabulary and English understanding , and use other prompting input (photos) as a base and/or build your own teaching library.

[ghosthost]

Thanks so much for that details response, Nimrod!

I'm getting some really great (for my purposes) results from the prompts I give--it's still a learning process, but my organic intelligence seems adaptable and capable of growth. But the current filters make anything erotic out of the question, so I'll take your advice and check out Stable Diffusion.

If I come up with anything really cool, I'll share it here.

[Shadowmuse Blown]

Awesome to hear the discussion between everyone who’s using AI.  Fascinating stuff, all.
sb_art69 at DA

[SamV]

Can I ask--what AI software is being used to produce fairly real looking mega-stacked babes? I've been experimenting with mid journey but they go berserk if there's even a tiny hint of sexuality in a prompt, like "big chest", etc. An exaggerated or naked breast would be impossible to create there, so...

There is no single perfect combination or software, although they all follow the same approach to generation.

Think of it as trying to tune in a radio station from a neighboring country. If the settings are off you will only get some Mexican radio and not the signal from Toronto.

It maters on:
The Software being used
The degree of control you have over the base algorithm (such as Stable Diffusion settings)
The limitations imposed by the software regarding NSFW content
The prompting seed TEXT or IMAGE (if any)
and most of all...
The library used to train the AI which not only imparts style but the ability for the AI to generate something. If it is not in the library the AI will struggle to "draw" the solution.

That is why you can often determine the library being used and not necessarily the software.
But that is only part of what it takes.

Even with the best of all the above running, if you can not provide a "conducive seed" then you are not going to get what you want/expect. Like trying to drive from New York to LA and starting by driving EAST.

My only suggestion it to work with the way you form your prompts, the manner in which you describe, and watch some examples. Most people run into problems because they are too direct with their seed text.

1. Avoid slang or jargon
2. Use more adjectives and fewer nouns
3. Describe the aspects and not the things
4. Use a deeper vocabulary while understand the subtle differences between synonyms

Realize that if the content is not in the teaching library the AI will be very hard pressed to create something because it really is not creating anything NEW, but rather the optimum solution that fits the prompts.

If you want total control, just download Stable Diffusion (free), find an interface app that you like, get a workable teaching library, and build up your vocabulary and English understanding , and use other prompting input (photos) as a base and/or build your own teaching library.

If I may also add something AI artists have said when I asked about how they get exactly what they want, is they will resort to using a graphics program like Photoshop. It also is useful in fixing AI screw-ups like "eight finger hands".  

BTW Nimrod, I really love your "Earth Goddess".  

[Nimrod]

BTW Nimrod, I really love your "Earth Goddess".

Thanks SamV - I will tell the her. Kithara was the base model, the AI just added the forest.

Kithara just smiled and says "thanks, it looks so much like me too! Although I'm not that blonde."

[MaxBigfoot]

 

[Shadowmuse Blown]

sb_art69 at DA

Another

[SamV]

BTW Nimrod, I really love your "Earth Goddess".

Thanks SamV - I will tell the her. Kithara was the base model, the AI just added the forest.

Kithara just smiled and says "Thanks, it looks so much like me too! Although I'm not that blonde."

Well Nimrod then I will take your ... I mean her word on it!  

[MaxBigfoot]

 

[Shadowmuse Blown]

From chuong1224 on DA

Sometimes fully clothed is really sexy

[SamV]

From chuong1224 on DA

Sometimes fully clothed is really sexy

No argument there, but what MaxBigfoot posted before your post, may say otherwise.