DPChallenge: A Digital Photography Contest You are not logged in. (log in or register
 

DPChallenge Forums >> Side Challenges and Tournaments >> Learning and Sharing Midjourney Images
Pages:  
Showing posts 1 - 25 of 74, (reverse)
AuthorThread
04/29/2023 11:29:29 AM · #1
This thread is about a journey into AI generated imagery, to learn and understand this new medium. If you are interested jump in to share your Midjourney images or to provide specific feedback on the images of this thread. Personally I find it fascinating and exciting but I am a technology addict :)
04/29/2023 11:30:36 AM · #2
I started learning Midjourney and tried various ideas. I used words from a site that generates prompts. I uploaded my own image:

as a starting point.

I used these words:
Photography::10000 insanely detailed and intricate, dreamy, hyper realistic, super detailed::10000
(::10000 means that these words are the most important)

and got these images:
. . .

That was not at all what I intended or expected, has anyone come across such strange styles in any kind of existing artwork? I tried Google Images but could not find any matches.
04/29/2023 12:09:42 PM · #3
That's wildly kinky, if that's the right word... Very cool in ts bizarre way...
04/29/2023 12:49:36 PM · #4
Those are gorgeous and weird. Not sure I'm ready to play with AI generated scenes yet.
04/29/2023 04:04:20 PM · #5
Is it possible that it used your browser cookies as data and introduced a photographic theme as a consequence?
04/29/2023 04:43:46 PM · #6
Originally posted by Paul:

Is it possible that it used your browser cookies as data and introduced a photographic theme as a consequence?


She wrote "Photography::10000" as part of her prompt
04/29/2023 05:17:45 PM · #7
In case anyone is not familiar, there is also a /describe function that will reverse engineer a prompt from a photo you upload and give you four descriptions that you can use as prompts by just clicking the corresponding numbers below the description output. Using Margaret's source image for example:



The clicking the generated prompts (unmodified, but you can also modify the prompt before committing), you get a set of four images for each:
1: 2: 3: 4:

NOTE: This feature might only be available on the premium plan. I can't remember.

ETA: And here's a modified version of prompt #1 just fer fun: ;-P



Message edited by author 2023-04-29 17:28:04.
04/29/2023 09:59:16 PM · #8
Originally posted by Paul:

Is it possible that it used your browser cookies as data and introduced a photographic theme as a consequence?
No, I am using Discord app so no browser is involved. I think now that Photography was interpreted as a general term while I wanted the output to be looking like a photograph. Learning how to prompt Midjourney is quite a challenge although results of mistakes can be fun :)
04/29/2023 10:09:56 PM · #9
Originally posted by Art Roflmao:

In case anyone is not familiar, there is also a /describe function that will reverse engineer a prompt from a photo you upload and give you four descriptions that you can use as prompts by just clicking the corresponding numbers below the description output. Using Margaret's source image for example:



The clicking the generated prompts (unmodified, but you can also modify the prompt before committing), you get a set of four images for each:
1: 2: 3: 4:
....


You got much better results from you descriptions than I did! I will try them later. One other thing I did was to put the /description result into ChatGPT to get some more flowery words and that gave me nicer images in result.

A very good resource that I got from Georges is this one:
https://promptomania.com/midjourney-prompt-builder/
That's where I got the prompts like "insanely detailed and intricate".
04/29/2023 11:00:16 PM · #10
Originally posted by MargaretNet:

You got much better results from you descriptions than I did!

Just to clarify, those weren't my desriptions. They were Midjourney's own /describe prompts.
04/30/2023 12:15:11 AM · #11
Difference between my description and the one generated by ChatGPT:

Mine:
"realistic photograph:: a man walking across the bridge towards an old windmill on the top of a small hill. The road has fences along with some lanterns. It is the middle of the night and an enormous moon is shining behind the windmill. Below the bridge there is a river and there is slight fog above it. On the right there are rocky hills with clouds on top of them. The bridge is made of stone, some landscape rocks have mauve flowers"

Resulting quad:


ChatGPT:
"realistic photograph:: a serene and picturesque scene set in the middle of the night. A man is seen walking across a stone bridge that spans a river, heading towards an old windmill perched atop a small hill in the distance. The road leading to the windmill is lined with fences and lanterns, casting a warm glow in the night. A colossal moon is shining brightly behind the windmill, illuminating the surrounding landscape. Above the river, there is a slight fog that adds a mystical ambiance to the scene. On the right side of the image, rocky hills rise with clouds lingering over their peaks, adding to the sense of mystery and grandeur. The bridge itself is made of stone, adding to the historic and rustic feel of the setting. Some landscape rocks along the bridge are adorned with mauve flowers, adding a touch of color to the scene."

Result:
04/30/2023 12:47:13 AM · #12
Another weird result:

Prompt: image + "hyperrealistic photograph::10000 --ar 3:2"

My image:


Midjourney:
04/30/2023 04:48:33 AM · #13
Originally posted by posthumous:

Originally posted by Paul:

Is it possible that it used your browser cookies as data and introduced a photographic theme as a consequence?


She wrote "Photography::10000" as part of her prompt


Doh! Not sure how my brain failed to discern that.
04/30/2023 08:48:09 AM · #14
Originally posted by MargaretNet:

I started learning Midjourney and tried various ideas. I used words from a site that generates prompts. I uploaded my own image:

as a starting point.

I used these words:
Photography::10000 insanely detailed and intricate, dreamy, hyper realistic, super detailed::10000
(::10000 means that these words are the most important)

and got these images:
. . .

That was not at all what I intended or expected, has anyone come across such strange styles in any kind of existing artwork? I tried Google Images but could not find any matches.


I love playing in Midjourney. Somehow I always get the dark images ... must be me :-).

I have never uploaded my own to generate something new. I think in your case it tried to "merge" the image you uploaded with the description, and took "photography" as meaning "camera" ... Not sure, does anybody really know how this works ? :-)
Did you see the Lindsey Adler youtube of how they took AI as inspiration and created the whole thing in "real life"? https://www.youtube.com/watch?v=9a5zleO_xrA

04/30/2023 09:10:45 AM · #15
Originally posted by kasaba:

...I have never uploaded my own to generate something new. I think in your case it tried to "merge" the image you uploaded with the description, and took "photography" as meaning "camera" ... Not sure, does anybody really know how this works ? :-)
Did you see the Lindsey Adler youtube of how they took AI as inspiration and created the whole thing in "real life"? https://www.youtube.com/watch?v=9a5zleO_xrA

Control of the output seems to be the problem, so far I have never got the same results with the same images and prompts, even with the lowest Stylize setting, but I am still very new to Midjourney, just 4 days old. I read some people illustrated storybooks with Midjourney images so they must achieved at least a consistent style.

I am not into fashion but I agree with the notion of this video - use Midjourney generated images for inspiration. In my case I have been looking for ideas for 3D scenes in Unreal Engine. If Midjourney could generate 3D meshes that would be even better! :)
04/30/2023 12:19:01 PM · #16
Just a heads-up: for those of you who are paid members and have a portfolio, it helps a LOT viewing your images if you drop the related ones into a new portfolio folder: that means viewers can scroll through in the same window/tab, instead of opening 4 windows/tabs to view 4 images.
04/30/2023 01:27:38 PM · #17
Attempting to replicate the features of a female model using prompts is a challenging task, especially when it comes to recreating the eyes, nose, and mouth accurately. However, generating hair seems to be a relatively simple process.
Ethnic traits are great. Using V5 improved this process significantly.
Long purple hair
Long purple hair
short purple hair
Latina Model
Belgian model
Scandinavian model
Apache Model
The prompt is simple and easy to understand
A beautiful girl with very long hair, dark purple hair color, in a dark purple dress, torso view, silver voluminous details. realistic expressive eyes, blurry lights in the background,--no bokeh style, real life, hyper detail, volumetric lighting, intricate, highly detailed
04/30/2023 03:39:04 PM · #18
.

Message edited by author 2023-05-01 16:36:49.
04/30/2023 03:40:00 PM · #19
Originally posted by posthumous:

Originally posted by GeorgesBogaert:


Apache Model


ah yes great example of RIRO

I'm talking about the datasets and the AI, not you Georges.
04/30/2023 06:45:17 PM · #20
I didn't know you could use your own images. If I could use it to make my images moody, etc, basically just advanced Photoshop techniques, that would be cool
04/30/2023 07:17:52 PM · #21
I've used it for illustrations with my grandson to put together our own interesting and unique versions of stories or characters he likes. We often get some crazy results, like this one that will likely get Thomas the Tank Engine canceled...


04/30/2023 07:53:57 PM · #22
I've used it to generate substitute stock images for clients websites. MJ v5 was much improved for this over v4 but still has issues with hands.
(prompt is in the description)

A couple others I've generated and used:
"a boston terrier"


"a pet-friendly hotel lobby"

I didn't think it looked enough like a hotel lobby, so I changed it to "a pet friendly hotel lobby with luggage" and it turned the furniture into luggage. LOL


Message edited by author 2023-04-30 19:59:21.
05/01/2023 03:24:14 AM · #23
Originally posted by vawendy:

I didn't know you could use your own images. If I could use it to make my images moody, etc, basically just advanced Photoshop techniques, that would be cool

This is what I was hoping that it will do. Unfortunately it always makes changes to the original greater than just editing. For example, I had a photo of a museum display with two kids (in Heidi village near Bad Ragaz), all results insisted on changing the kids! They were much worse than the original. I also hoped that I will be able to use /blend command for combining two images for the Extended FS, that did not work either, the results always involved something that was not in the originals.
05/01/2023 07:11:24 AM · #24
Originally posted by Art Roflmao:

I've used it to generate substitute stock images for clients websites. MJ v5 was much improved for this over v4 but still has issues with hands.
(prompt is in the description)...

I ran your prompt out of curiosity and got some weird stuff too:


On #1 the small kid has too big hand, #2 - the girl seems to have very long fingers, the guy on #3 looks very much like young Federer, on #4 the knee of a woman looks like she was working out too much. I guess majority of people would not notice these details when they use the pet website.

Message edited by author 2023-05-01 13:48:20.
05/01/2023 10:14:53 AM · #25
I have a show this weekend, so I can't play around with it. But can you give the basic information of what I need to do to get started, so I can save some digging, but just jump in when I have time?

Is it web-based? is it software I need to purchase? (I'd never heard of it before now. Only played around with Dall-e, and that was just for a couple of tries.

Thanks!
Pages:  
Current Server Time: 03/28/2024 12:40:47 PM

Please log in or register to post to the forums.


Home - Challenges - Community - League - Photos - Cameras - Lenses - Learn - Prints! - Help - Terms of Use - Privacy - Top ^
DPChallenge, and website content and design, Copyright © 2001-2024 Challenging Technologies, LLC.
All digital photo copyrights belong to the photographers and may not be used without permission.
Current Server Time: 03/28/2024 12:40:47 PM EDT.