NightCafe Studio Has Upped It's Game!

[RedKittieKat]

When I first tried NightCafe many months ago, I was not impressed. Especially after my Midjourney experiments. However, they have recently updated their nuts and bolts with something called SDXL 0.9.

It’s a big upgrade. The difference is quite stunning. Sans hands and legs … I’ll be completely amazed once any AI learns how to do that

These are nearly the same prompts. They both included sunset, ethereal, gentle, delicate, flowing lace gowns. The first also included Woman standing alone and the second is Mother holding baby (I couldn’t use child it’s a banned word)

Here is a sample of many months ago - complete garbage.

And today:

It made me really miss Midjourney I just haven’t decided to bite the bullet and get a subscription yet. But, NightCafe was mighty impressive this morning. Of course those that play with this all day every day are having some pretty amazing results.

[PrintDriver]

It’s a sad world where ‘child’ is a banned word in an art AI.
Sad indeed.

But the difference in the generated art is amazing!

[RedKittieKat]

I was completely baffled. My prompt came back as containing banned words. I had to go searching because NC doesn’t give a list of what words it bans. You have to guess

I couldn’t imagine what I typed that was triggering this. I finally found a post on Reddit. You can’t use Child or Kid in any prompt. Baby, however worked fine.

[RedKittieKat]

My goggle kitty is cracking me up

[Just-B]

I like it!

[RedKittieKat]

Won myself 20 credits to keep playing another day

[PrintDriver]

AI does steampunk art very well!
I really have to start trying this stuff. Still loath to learning the prompts necessary though.

[RedKittieKat]

My prompts are always fairly simple. I see some that are two paragraphs long. That will never be me lol For this Robot challenge it was ‘‘Horse in motion, Robotic, Gold and black metal, ornate, futuristic’’. and then I picked their newest generator and the vibrant color theme. Besides, I don’t understand half of the descriptive terms some people use. It’s a secret code I think

[RedKittieKat]

Pirate Cats in Space … and a kitten

[PrintDriver]

Permission to steal that second one for my desktop?

[RedKittieKat]

Sure! That one earned me a few more tokens to keep playing the challenges

They are fun because it makes me think of prompts I never would have thought of.

[PrintDriver]

Thanks!
I love it overall but the cap, with what looks like a cat skull crunching a bone in half…if this were not an AI I’d give the artist serious kudos for that!

[RedKittieKat]

A rose by any other name …

[RedKittieKat]

I’ve actually engaged in the community on Nightcafe … eek … I know
Boy … some just seem so young. They are very worried about likes and if people approve of what they did. If they don’t get recognition they get very defensive. Sometimes to the point the Mods have to step in and tell them to shut it or be gone. There are also those who work very hard to get around the sensors so a boob shows up. They get kicked fast. Other than that, it’s fairly tame and drama free.

I’ve been learning a few things. In addition to NC’s daily competitions, there are nightly ones as well, hosted by the Newbie Mods to help work out prompts and share ideas (although some newbies refuse to share).

Here are a few things I’ve had fun with:

A Gentlemans Portrait. He ended up looking like Cary Grant LOL and the majority took that to mean a Viking covered in tats. I think it’s a generational thing

Party Otter:

Planets & Galaxy:

Stained Glass:

Joy:

Underwater:

[Just-B]

These are great!

I don’t know how you’re creating them. I played around with NightCafe for around an hour yesterday and couldn’t get it to build anything close to as nice as these. For example, I asked it to draw a profile of a young woman crossing a busy city street, and it created a woman with three legs and two torsos emerging from her skirt.

[RedKittieKat]

Thanks Just-B!!

I have found that you get the best results if you can be as descriptive as possible with the least amount of words. But, I see some with prompts as long as my arm. Honestly, you are only seeing what I think are my best as well. I was trying to make some shoes and kept ending up with 3 heels and one leg or 2 legs and one shoe. I gave up. Sometimes the AI just doesn’t get what my intent is.

Here are the settings I always use:

SDXL 0.9 only. The rest are pretty much useless as far as I’m concerned. They have just come out with version 1.0 which is for PRO only right now.

For style I only use Hyperralistic or Vibrant. I use Artist Portrait if I want something softer and flowing.

I also change one or two descriptive terms at a time.

Here is what you said you were looking for: '‘profile of a young woman crossing a busy city street’'

With those two settings and this exact wording this is what I got:

Ok … not right … it’s thinking I just want a side shot of her face. So I changed the prompt slightly to indicate I wanted legs.

"Young woman walking across a busy city street"

This is what I got:

It’s getting there … but still not getting that I want a full view. So …

"Full body of Young woman walking across a busy city street’'

Getting there but I still need to see a profile view … so …

’‘Side view, full body of young woman walking across a busy city street’'

Success!

That’s how I try to get the AI to do what I want. But, bear in mind … sometimes it just doesn’t cooperate I see some on there with thousands of created images and only a handful published by comparison. So you know there are a lot that don’t turn out for most the first go.

Don’t get frustrated … just change course if needed.

[Just-B]

I didn’t get anything even remotely that good — despite several tries. I think I might have been too specific in my requests. The sentence I posted here was just a summary of my more elaborate string of words. I’ll try boiling things down to less specific requests that won’t confuse it.

[RedKittieKat]

Yep … just keep trying. You will get a good feeling of what will work and what won’t in no time.

[Just-B]

Using the same simple prompt, “Young woman walking across a busy city street”, here’s what I got: a four-armed woman.

The AI imaging apps don’t seem to like me.

[Smurf2]

Ripley?