AI Image Generators - Any Good Ones?

[mopek1]

I'm looking mostly for mainstream images to create for blogs I want to start.

Any good free or paid ones you recommend?

[just a punk]

Stable Diffusion XL generates absolutely cool images for my autoblogs with these very simple prompts:

1. A highly realistic image for a blog article on topic of "%post_title%" - works best for "photographic" style preset.

2. A cool image for a blog article on topic of "%post_title%" - this one is for "anime", "digital art" and "neon punk" presets.



The Stability AI API is very cheap, like $10 for thousand of generated images.

P.S. That's for mainstream of course...

[Mr Pheer]

Originally Posted by just a punk

Stable Diffusion XL generates absolutely cool images for my autoblogs with these very simple prompts:

1. A highly realistic image for a blog article on topic of "%post_title%" - works best for "photographic" style preset.

2. A cool image for a blog article on topic of "%post_title%" - this one is for "anime", "digital art" and "neon punk" presets.



The Stability AI API is very cheap, like $10 for thousand of generated images.

P.S. That's for mainstream of course...

Why did you change your name? I had no idea who you were until you just posted some of the same stuff I'm using..

[just a punk]

Originally Posted by Mr Pheer

Why did you change your name? I had no idea who you were until you just posted some of the same stuff I'm using..

Because I'm a punk

By the way, have you noticed how good Stable Diffusion XL is at generating images for blog posts based on their titles only? DALL-3 is also not that bad as the 2nd generation, but it is about 40 times more expensive...

[Mr Pheer]

Originally Posted by just a punk

Because I'm a punk

By the way, have you noticed how good Stable Diffusion XL is at generating images for blog posts based on their titles only? DALL-3 is also not that bad as the 2nd generation, but it is about 40 times more expensive...

Well that's because SDXL is pretty good, I also run it on my local machine when I'm not playing BG3

My favorite and mostly used one is still Midjourney though. But not for autoblogging stuff, obviously.

[mopek1]

Originally Posted by just a punk

Stable Diffusion XL

Thanks. I'll check it out. And thanks for the prompts too. The price is pretty good. If you make a mistake though, does that count as an image towards your 1000 for $10?

[mopek1]

Originally Posted by Mr Pheer

My favorite and mostly used one is still Midjourney though.

This is the only one I've heard of so far. Another parent at my kid's school told me about it. I wanted to see if anyone mentioned it in this thread. Which membership option do you have?

[just a punk]

Originally Posted by mopek1

Thanks. I'll check it out. And thanks for the prompts too. The price is pretty good. If you make a mistake though, does that count as an image towards your 1000 for $10?

To be honest, I didn't get your question. Could you please rephrase it?

[just a punk]

Let's assume you have an automotive blog running on the CyberSEO Pro plugin, which automatically posts new articles daily. These articles are generated by OpenAI GPT based on a list of article titles you've provided. Suppose you also have a cooking recipes blog and a blog dedicated to pets. You can find detailed instructions on how to automatically generate unique articles with HTML formatting for these blogs on the plugin's website, which includes documentation and numerous articles with detailed guidance.

However, you'll also need images for these articles. The plugin can import images from Google Image search or Pixabay stock, but these won't be unique. For this reason, I recommend using built-in tools like DALL-E 3 and Stable Diffusion XL. These tools can generate excellent images for your articles based solely on the topic titles.

Let's take, for example, article titles from both of your blogs: "Roasted Garlic Pasta" from the cooking blog and "Puppy Biting: What's Normal, What's Not, and How to Curb It" from the pet blog. You would give DALL-E 3 and Stable Diffusion XL a simple task like this (in the CyberSEO Pro settings, the shortcode %post_title% will automatically be replaced with the title of the current post being generated):

"A highly realistic image for a blog article on the topic of '%post_title%'"

As a result, the plugin will add images like these to your articles:

DALL-E 3 for "Roasted Garlic Pasta":


DALL-E 3 for "Puppy Biting: What's Normal, What's Not, and How to Curb It":


Stable Diffusion XL for "Roasted Garlic Pasta":


Stable Diffusion XL for "Puppy Biting: What's Normal, What's Not, and How to Curb It":


I think the image quality is ok, suitable for an autoblog that runs on full autopilot and does not require any manual input.

This is not Midjourney, but DALL-E 3 and Stable Diffusion, which are integrated with the plugin via REST API. And these images are not handpicked 1 out of 4. They are the first ones autogenerated for given prompts.

[mopek1]

Originally Posted by just a punk

Let's assume you have an automotive blog running on the CyberSEO Pro plugin, which automatically posts new articles daily. These articles are generated by OpenAI GPT based on a list of article titles you've provided. Suppose you also have a cooking recipes blog and a blog dedicated to pets. You can find detailed instructions on how to automatically generate unique articles with HTML formatting for these blogs on the plugin's website, which includes documentation and numerous articles with detailed guidance.

However, you'll also need images for these articles. The plugin can import images from Google Image search or Pixabay stock, but these won't be unique. For this reason, I recommend using built-in tools like DALL-E 3 and Stable Diffusion XL. These tools can generate excellent images for your articles based solely on the topic titles.

Let's take, for example, article titles from both of your blogs: "Roasted Garlic Pasta" from the cooking blog and "Puppy Biting: What's Normal, What's Not, and How to Curb It" from the pet blog. You would give DALL-E 3 and Stable Diffusion XL a simple task like this (in the CyberSEO Pro settings, the shortcode %post_title% will automatically be replaced with the title of the current post being generated):

"A highly realistic image for a blog article on the topic of '%post_title%'"

As a result, the plugin will add images like these to your articles:

DALL-E 3 for "Roasted Garlic Pasta":


DALL-E 3 for "Puppy Biting: What's Normal, What's Not, and How to Curb It":


Stable Diffusion XL for "Roasted Garlic Pasta":


Stable Diffusion XL for "Puppy Biting: What's Normal, What's Not, and How to Curb It":


I think the image quality is ok, suitable for an autoblog that runs on full autopilot and does not require any manual input.

This is not Midjourney, but DALL-E 3 and Stable Diffusion, which are integrated with the plugin via REST API. And these images are not handpicked 1 out of 4. They are the first ones autogenerated for given prompts.

Yeah, I get it now. Thanks for the reply. They aren't bad for 'realistic' and like you said, for autoblogging that's all you need really.

[2MuchMark]

Check this out



You can guess most of the prompts, but I included some extras such as mood, color grading, time of day, etc. I was blown away and it makes me feel really sad to look at it, as it should.

In case you didn't know, you can even specify the type of lens and camera that can be used. Be sure to use words like "Photorealistic" and "Ray Tracing" too for some even better results.

[mopek1]

Originally Posted by 2MuchMark

Check this out

.

Really nice! What generator was that made with?

[2MuchMark]

Originally Posted by mopek1

Really nice! What generator was that made with?

Dall - E.

[2MuchMark]

Originally Posted by mopek1

Really nice! What generator was that made with?

I also made this one



But since I am more of a cat person, the "Old Man and the Kitty alone and ignored in the Rain" hits harder to me. The only problem with both images is that neither the cat nor the dog are wet, despite being in the rain.

[just a punk]

Originally Posted by 2MuchMark

Check this out



You can guess most of the prompts, but I included some extras such as mood, color grading, time of day, etc. I was blown away and it makes me feel really sad to look at it, as it should.

In case you didn't know, you can even specify the type of lens and camera that can be used. Be sure to use words like "Photorealistic" too for some even better results.

Could you provide your prompt? I'd like to see how it works with Stable Diffusion. DALL-3 is cool, but its API is too costly...

[just a punk]

Stable Diffusion XL with some standard style presets.

Photographic #1:



Photographic #2:



The same with the "anime" preset:



Neon punk:



Analog film:



P.S. As I mentioned above DALL-E 3 is also cool and it is also supported, but it's too costly...

[Mr Pheer]

There is some that I just do for fun on facebook. Not anything to make money with, just a creative outlet. Like and follow if you want.

https://facebook.com/mrpheer

There's just something funny about spongebob running around with a gun...

[Mr Pheer]

Originally Posted by mopek1

This is the only one I've heard of so far. Another parent at my kid's school told me about it. I wanted to see if anyone mentioned it in this thread. Which membership option do you have?

I use the $30 option but sometimes I upgrade to the $60 one for more turbo hours. Seems like I'll be moving to that level as a baseline now that the midjourney tuner works and I use it quite a bit.

[Mr Pheer]

Originally Posted by just a punk

Photographic #2:

I pasted it in Midjourney and used the describe function, got 4 new prompts and ran them. Then I picked my favorite, it was zoomed in real close so I zoomed it out 2x and let Midjourney do the inpainting... and here we go.




an old man and cat sit on a bench in the rain, in the style of movie still, lightbox, orange, nikon d850, hasselblad 500c/m, inventive character designs --ar 16:9 --style raw --s 750



Upscale Img 2



Zoom Out 2x



Here is another one I liked



Cat was lonely after the old man died of pneumonia from sitting in the cold rain.



These are just for fun... not the way to do it for autoblogging.

[fris]

midjourney does the best ones

[just a punk]

Originally Posted by fris

midjourney does the best ones

That's for sure. The only problem - it has no official API...

[JesseQuinn]

this thread is awesome, y'all made some cool shit


(now lonely kitty needs a home)

[Klen]

Originally Posted by fris

midjourney does the best ones

It depend what you want to generate
For example, for food dall e is best, with only downside of generated lighting, which kind make it obvious how is AI generated picture, while midjourney is best for generating real human pictures.

[mopek1]

Originally Posted by 2MuchMark

Check this out



You can guess most of the prompts,

Question. What's if I wanted to generate 10 types of images like this one? But with a different looking old man, different, looking pet, different looking city etc., for each image?

What kinds of prompts would I use?

If you ask the generator to create an old man in the rain, will it always create that same looking man? Or will it randomly create an old man that is different in these ways: height, weight, age, skin tone, build, look ...

Same question about setting or animals, or any other feature you ask it to create.

So if my blog post was about fashion, and I wanted 10 different images of a fashionable lady, would it create the same lady 10 times ?

[just a punk]

Originally Posted by mopek1

Question. What's if I wanted to generate 10 types of images like this one? But with a different looking old man, different, looking pet, different looking city etc., for each image?

What kinds of prompts would I use?

If you ask the generator to create an old man in the rain, will it always create that same looking man? Or will it randomly create an old man that is different in these ways: height, weight, age, skin tone, build, look ...

Same question about setting or animals, or any other feature you ask it to create.

So if my blog post was about fashion, and I wanted 10 different images of a fashionable lady, would it create the same lady 10 times ?

In CyberSEO Pro you can use Spintax in your prompts. E.g.:

{an old|a seeping|a drunk|a crazy} {man|woman} and a {white|black|grey|red} {cat|horse|hamster} sit on a {bench|rock} {in the {rain|park}|on the Moon}

So when the plugin will generate a new post, your image prompt will be different

Also, don't forget to use the %post_title% shortcode in your prompts, it will be replaced with the actual post title each time a new post is created. It will tell the AI image generation engine what exactly to draw there.

[mopek1]

Originally Posted by just a punk

In CyberSEO Pro you can use Spintax in your prompts. E.g.:

{an old|a seeping|a drunk|a crazy} {man|woman} and a {white|black|grey|red} {cat|horse|hamster} sit on a {bench|rock} {in the {rain|park}|on the Moon}

.

I bought that awhile back but was overwhelmed once I installed it. I didn't see a quick start tutorial and after looking through it for an hour gave up. Maybe I'll give it another go.

[2MuchMark]

Originally Posted by Mr Pheer

I pasted it in Midjourney and used the describe function, got 4 new prompts and ran them. Then I picked my favorite, it was zoomed in real close so I zoomed it out 2x and let Midjourney do the inpainting... and here we go.




an old man and cat sit on a bench in the rain, in the style of movie still, lightbox, orange, nikon d850, hasselblad 500c/m, inventive character designs --ar 16:9 --style raw --s 750



Upscale Img 2



Zoom Out 2x



Here is another one I liked



Cat was lonely after the old man died of pneumonia from sitting in the cold rain.


.

Well holy shit I think I want to kill myself now... those are amazing renders. The problem I have with alot of the AI images is that they are very boring unless you inject emotion into the problems. "Sad", "Lonely" work... even "Ignored" as in "the people walking by are all ignoring the old man and cat". You can also tell the DALL - E a history of the person or people in the image. "The old man had a rough life and slowly lost everything including his family and his home. This caused him to smoke and drink alot and his hardship should show in his face".

[2MuchMark]

Originally Posted by mopek1

Question. What's if I wanted to generate 10 types of images like this one? But with a different looking old man, different, looking pet, different looking city etc., for each image?

What kinds of prompts would I use?

If you ask the generator to create an old man in the rain, will it always create that same looking man? Or will it randomly create an old man that is different in these ways: height, weight, age, skin tone, build, look ...

Same question about setting or animals, or any other feature you ask it to create.

So if my blog post was about fashion, and I wanted 10 different images of a fashionable lady, would it create the same lady 10 times ?

In my experience it always creates a different looking character however I think you can now use restraints. For example if you give that man a name, and then tell the AI to make him again but wearing different clothing for example, it will keep the same facial features and hair, and change everything else. I have not tried this personally but I heard that it works.

[fris]

Midjourney, ChatGPT, DALL·E, Stable Diffusion prompts on here.

https://promptbase.com/

can buy and sell your prompts.

also

https://prompthero.com/

is good

[just a punk]

Originally Posted by mopek1

I bought that awhile back but was overwhelmed once I installed it. I didn't see a quick start tutorial and after looking through it for an hour gave up. Maybe I'll give it another go.

Here they are:

1. How to add an RSS feed and generate additional content for each imported post. How to completely rewrite the posts. How to translate them into other languages (all that in a 9-minute video):




2. How to generate new posts from scratch (no-source autoblogging):




3. Advanced no-source autoblogging (generating long and rich articles on full autopilot):

[Mr Pheer]

Originally Posted by 2MuchMark

Well holy shit I think I want to kill myself now

Ok

[Mr Pheer]

Originally Posted by mopek1

Question. What's if I wanted to generate 10 types of images like this one? But with a different looking old man, different, looking pet, different looking city etc., for each image?

What kinds of prompts would I use?

If you ask the generator to create an old man in the rain, will it always create that same looking man? Or will it randomly create an old man that is different in these ways: height, weight, age, skin tone, build, look ...

Same question about setting or animals, or any other feature you ask it to create.

So if my blog post was about fashion, and I wanted 10 different images of a fashionable lady, would it create the same lady 10 times ?

I don't know about DallE, but both Midjourney and Stable Diffusion will both let you use reference photos to give it an idea. And Midjourney has a tuner, you can really get damn close to what you want with consistent characters in different scenes. I was doing it the hard way back in March but now it's so simple. Simple as in, not for an idiot but an experienced webmaster shouldn't have a hard time with it.

[Mr Pheer]

Originally Posted by mopek1

I bought that awhile back but was overwhelmed once I installed it. I didn't see a quick start tutorial and after looking through it for an hour gave up. Maybe I'll give it another go.

He used ChatGPT and wrote pretty good instructions and made videos, still overwhelming at times but easier to understand and use it.

[2MuchMark]

Originally Posted by Mr Pheer

Ok

[2MuchMark]

Originally Posted by Mr Pheer

I don't know about DallE, but both Midjourney and Stable Diffusion will both let you use reference photos to give it an idea.

I tried using reference photos with DallE but it wouldn't let me. They also have some kind of copyright thing in place. I tried making images with "Captain Kirk" but it wouldn't work. I then tried "A space ship that looks like the Starship Enterprise" but that too was blocked.

[2MuchMark]

Originally Posted by Mr Pheer

an old man and cat sit on a bench in the rain, in the style of movie still, lightbox, orange, nikon d850, hasselblad 500c/m, inventive character designs --ar 16:9 --style raw --s 750

I tried using your prompt above in Dall-E and got this:



It is clearly trolling me. Bastard. !

[Mr Pheer]

Originally Posted by JesseQuinn

this thread is awesome, y'all made some cool shit


(now lonely kitty needs a home)

You get this cat. And 6 fingers.



Enjoy the cat.

[mopek1]

Originally Posted by Mr Pheer

He used ChatGPT and wrote pretty good instructions and made videos, still overwhelming at times but easier to understand and use it.

Thanks, I'll give it another look when I get the chance.

[mopek1]

Originally Posted by Mr Pheer

both Midjourney and Stable Diffusion will both let you use reference photos to give it an idea. .

Thanks. I just tried it. The first photo I used just for practice was of a Pilates woman doing a stretch and midjourney wouldn't allow it because they said she was shirtless

She had on a small tank top that most women wear to exercise classes, skimpy, bit very common. It seems very sensitive to too much skin showing. I've seen people's art with anime superhero type girls with skimpy bikinis and such.

Oh well, back to practicing, but it's really neat so far.

[mopek1]

Originally Posted by just a punk

Here they are:

Awesome, thanks!

[Mr Pheer]

Originally Posted by 2MuchMark

I tried using your prompt above in Dall-E and got this:



It is clearly trolling me. Bastard. !

Yeah it's trolling you

But is it? You mentioned a camera, it gave you a camera

[Alex CZ]

Originally Posted by 2MuchMark

I tried using your prompt above in Dall-E and got this:



It is clearly trolling me. Bastard. !

Is it part of chatGPT ? If I have premium membership with Chat is this tool free? All I see is "buy credits".

[Mr Pheer]

Originally Posted by 2MuchMark

I tried using reference photos with DallE but it wouldn't let me. They also have some kind of copyright thing in place. I tried making images with "Captain Kirk" but it wouldn't work. I then tried "A space ship that looks like the Starship Enterprise" but that too was blocked.

Looks like Midjourney doesn't want to do Star Trek characters without a little tricky prompting... I did get one in the "style" of Kirk on the Enterprise with some cute chick... he was always with some cute girls

[Mr Pheer]

Originally Posted by Alex CZ

Is it part of chatGPT ? If I have premium membership with Chat is this tool free? All I see is "buy credits".

Dall-E is part of ChatGPT. Although I have a paid account and also have API access. It even works if you can't spell image. First time I've tried it with the regular command line.

[Mr Pheer]

Heh, found the limit of the filter. It won't do it as a photo, but it will do Star Trek all day long as a drawing.

[Mr Pheer]

Originally Posted by Alex CZ

Is it part of chatGPT ? If I have premium membership with Chat is this tool free? All I see is "buy credits".

Use this setting

[JesseQuinn]

Originally Posted by Mr Pheer

You get this cat. And 6 fingers.



Enjoy the cat.

lol, I shall!

with 6 fingers, all the better for petting Ms Wilhelmina Smyth. she'll be a very happy Carib-cat


it's interesting though, that ai sucks with hands. in offline art those are the hardest to draw/paint/sculpt. way more than feet

funny that computer systems grapple with the same. you can always tell ai by the hands. someone needs to train the programs on some Picasso blue period and Rodin


thank you for the cat. I found a home for lonely kitty with a friend, so Mark please don't kill yourself



have no interest in creating ai stuff myself anymore (found it fun for a minute, then lost interest), but find convos like this fascinating just seeing what y'all make and how. again, cool as fuck thread

[AdultB2B]

Check out deepfake.com

[JesseQuinn]

^^^

that has nothing to do with the OP's question (and from there just started having fun with lots of super-insightful info, kinda how good peeps roll), but great contribution from you I guess. did you even read the thread?


Cyber/Punk wish you well with your solid product that can do a lot

[AdultB2B]

Originally Posted by JesseQuinn

^^^

that has nothing to do with the OP's question (and from there just started having fun with lots of super-insightful info), but great contribution from you I guess. did you even read the thread?


Cyber/Punk wish you well with your solid product that can do a lot

I guess I read the headline.