Media Arts and Technology

Project 3: MidJourney

ASSIGNMENT
For this assignment we are continuing with MidJourney but introduce img2img, a technique by which one or more images are used to generate a new image. Additionally, there is also the added function of a text prompt.

Recommendations for the assignment - Try combining 2,3,4, etc. images as your image prompts. Post a minimum of 2 images per multiple image prompts studies. For instance, you might want to test which combination is of interest to you, 2,3,4,etc. If any of these seem to give interesting results, then stay with the combination and test diferent ways to combine the source images, as the order of which image is listed first in the prompt may determine what the results will be like.

What may be other conceptual questions in using multiple images? Try to go beyond conventional results with the intent to test what may be limits of how MidJourney works. Will MidJourney produce total noise? Will it organize the image space based on visual elements from each of the image prompts? ...Will it try to impose a style, or aesthetic? In the end, your goal is to arrive at results that define your aesthetic interests, not the original images', nor the MidJourney aesthetic.

INSTRUCTIONS
Instructions for how to multiple images as prompts are posted at the MidJourney Docs: https://docs.midjourney.com/docs/image-prompts
Instructions provide how to find an image's web link.
The image weight parameter --iw provides a way to control the importance of the image in relation to the text prompt
The MULTI_PROMPT "::" function described at https://docs.midjourney.com/docs/multi-prompts allows for giving different weights to individual words in your text prompt.
There are a number of websites that discuss these techniques, for instance: https://graphicsgurl.com/midjourney-pro ... _article=1

In accordance with this week's assignment, I undertook the task of amalgamating three distinct visual elements: the seminal proto-impressionist masterpiece "The Scream" (1893) by the Norwegian artist Edvard Munch, Henri Cartier-Bresson's iconic 1946 portrait of philosopher Jean-Paul Sartre on a bridge over the Seine in Paris, and my own image of a solitary figure on a bridge under a cloudy sky, crafted previously on Midjourney. The underlying concept that drove the selection of these three images was the shared compositional elements they possessed. Each of them featured a solitary character positioned on a bridge-like structure that receded into the background, emphasizing a one-point geometric depth perspective. Additionally, they all drew attention to the sky in the background and presented ample headspace, establishing a common ground for intriguing compositional intersections. The experiment aimed to explore how three distinct images, varying in aesthetic style and originating from diverse media, could harmonize and to determine which aesthetic style would dominate. Multiple combinations were tried, along with different weighting values, to assess how Midjourney could produce complex and captivating amalgamations, thereby pushing the boundaries of AI image creation on Midjourney. I have shared multiple images from each combination below.


Image series 1








Prompt for 1: https://s.mj.run/8e_BXxx8SYI:: 0.5 ::https://cdn.discordapp.com/attachments/ ... 30d7b9bd15&:: https://s.mj.run/nVsYLfePa4o as a picture-postcard in black and white Kodak TriX 400 ISO, use dramatic lighting and shadows --ar 16:9 --c 10 --s 250 --style raw - @MAT 255 (fast)

For the first series of combinations, I gave more weight to my own image, assigning it a value of 0.5 times that of Munch's work, followed by Cartier-Bresson's portrait. The results highlight the headspace as the central dramatic element. Notably, the surreal, tunnel-like structure in the clouds deviated from my original prompt, surpassing my expectations in AI image generation. However, due to the weightage values in the prompt, the outcomes closely resembled my original image.

Image series 2









Prompt for 2: https://s.mj.run/3s6dGWdLvvY:: 2::https://cdn.discordapp.com/attachments/ ... d411087c38&:: 0.5::https://cdn.discordapp.com/attachments/ ... 70ba349d17& as a picture-postcard in black and white Kodak TriX 400 ISO, use dramatic lighting and shadows --ar 16:9 --c 10 --s 250 --style raw - @MAT 255 (fast)

In the second series, I initiated the prompt with Munch's painting, giving it double the weightage compared to Cartier-Bresson's photograph, which, in turn, had half the weightage of my image. The result featured a ghastly face reminiscent of Munch's figure in "The Scream" on the bridge. Midjourney seemed to have followed the prompt by organizing the image space based on visual elements according to the assigned weightage values. It attempted to blend Munch's painting style with the photographic realism I had requested, yielding an intriguing hybrid. In this case, Midjourney met me halfway in realizing my aesthetic interest, with the resulting image falling between Midjourney's imposed aesthetic, mine, and the original images.


Image series 3










Image series 4









Image series 5








Prompt for 3: https://s.mj.run/nVsYLfePa4o:: 1.5::https://cdn.discordapp.com/attachments/ ... 30d7b9bd15&:: 0.5::https://cdn.discordapp.com/attachments/ ... 70ba349d17& as a picture-postcard in black and white Kodak TriX 400 ISO, use dramatic lighting and shadows --ar 16:9 --c 10 --s 250 --style raw - @MAT 255 (fast)

Prompt for 4: https://s.mj.run/nVsYLfePa4o:: 0.5::https://cdn.discordapp.com/attachments/ ... 70ba349d17&:: 1.5::https://cdn.discordapp.com/attachments/ ... 30d7b9bd15& as a picture-postcard in black and white Kodak TriX 400 ISO, use dramatic lighting and shadows --ar 16:9 --c 10 --s 250 --style raw - @MAT 255 (fast)

Prompt for 5: https://s.mj.run/nVsYLfePa4o:: 0.5::https://cdn.discordapp.com/attachments/ ... 70ba349d17&:: 2::https://cdn.discordapp.com/attachments/ ... 30d7b9bd15& as a picture-postcard in black and white Kodak TriX 400 ISO, use dramatic lighting and shadows --ar 16:9 --c 10 --s 250 --style raw - @MAT 255 (fast)

In all three series, Munch's style took precedence over other styles, likely because it was listed first and assigned the highest weightage value. Even when my image sometimes carried more weight than Cartier-Bresson's portrait, the resulting images bore little resemblance to my original image, often resembling Sartre, complete with his distinctive glasses and hairstyle. These experiments did not conclusively align with my aesthetic preferences, as they predominantly borrowed from the artists' works I had used. Nonetheless, they provided a fascinating exploration of how AI integrated these styles to create a photo-impressionist, cartoonish depiction of a Sartre-like figure.


Image series 6








Prompt for 6: https://s.mj.run/8e_BXxx8SYI:: 0.5::https://cdn.discordapp.com/attachments/ ... d411087c38&:: 1.5:: https://s.mj.run/3s6dGWdLvvY as a picture-postcard in black and white Kodak TriX 400 ISO, use dramatic lighting and shadows --ar 16:9 --c 10 --s 250 --style raw - @MAT 255 (fast)

Here, Cartier-Bresson's photograph was mentioned first and assigned the highest weightage value, followed by Munch's work, with my image holding 1.5 times the weightage of Munch's painting. The resulting dystonic aesthetic style closely matched my intentions, making it a more accurate representation of my aesthetic preferences.


Image series 7








Prompt for 7: https://s.mj.run/3s6dGWdLvvY:: 2:: https://s.mj.run/nVsYLfePa4o:: 1.5::https://cdn.discordapp.com/attachments/ ... 70ba349d17& as a picture-postcard in black and white Kodak TriX 400 ISO, use dramatic lighting and shadows --ar 16:9 --c 10 --s 250 --style raw - @MAT 255 (fast)

In my final series of images, I mentioned my own image first in the prompt but assigned more weight to the other two. Surprisingly, Munch's painting, listed last, held the highest weightage value of 1.5. The outcome leaned more towards my image than the other artists' styles. Notably, the background exhibited architectural elements reminiscent of a European town, possibly influenced by giving Cartier-Bresson's work a higher weightage than mine. Although the AI adhered to my original image, I expected a novel aesthetic style that would align with my vision but differ from my prompt image. Instead, the result was a blend of the aesthetic styles used in the prompt.

I must acknowledge that I had requested the resulting images to resemble picture-postcards made from Kodak Trix-400 black and white film stock. Although some images appeared desaturated, none were entirely black and white, and none resembled picture-postcards. It's evident that I need to refine my prompts and use specific commands to prioritise the desired aesthetic style. Additionally, experimenting with more dramatic weighting values, such as 3 or even 6, and combining more images in future endeavours may help explore the limits of AI image creation on Midjourney. In addition to this weeks assignment, I have also tried animating some of the images with AI engine provided by Pika Labs. Below are some examples with prompts:

Video Prompt 1: imagine rotating tunnel man walking --ar 16:9
https://discord.com/channels/1023109610 ... 6210924645

Video Prompt 2: monster taking to a man
https://discord.com/channels/1023109610 ... 6297321592

Video Prompt 3: turbulent river flowing in the background
https://discord.com/channels/1023109610 ... 7474664498

Video prompt 4: heavy rain, water flowing
https://discord.com/channels/1023109610 ... 7181467728

Video prompt 5: heavy rain, man turning his head
https://discord.com/channels/1023109610 ... 5370132601

(1-5) The beginning series of images captured what I expected from Midjourney to default to in terms of over-stylization and imposition of content. Additionally, I was struggling to get the software to accept my image formats. The second set of images took a strange turn when adding two images. It still heavily imposed a certain aesthetic and overly detailed and too-dimensional version of what I was looking for. The characters that appear within the scenes also took a strange route as none of the images I uploaded felt visually cohesive with this in terms of subject. There is a color consistency that Midjourney did successfully pick up on and general border/foreground repetition based on the images uploaded. 
1. https://s.mj.run/T-iOLQw8KP0 :: dr.suess cartoon comic book scene, shark, fun scene, underwater, sea of fish --ar 16:9 --stylize 750 --style raw - @MAT 255 (fast)

2. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/nKdNHnvmyME :: dr.suess cartoon comic book scene, fun scene, underwater, sea of fish --ar 16:9 --stylize 750 --style raw - @MAT 255 (fast)

3. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: dr.suess cartoon comic book scene, shark, fun scene, underwater, sea of fish --ar 16:9 --stylize 750

4. Theodor Seuss Geisel cartoon comic book scene, fun, underwater, sea of fish --ar 16:9 --stylize 750

5. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats --ar 16:9 --stylize 50

(6-8) For this a new set of images was uploaded. Within that, there was a comic book strip featuring Garfield and a scene from Dr. Suess. At the beginning of this merge, the styles are extremely apparent and the effort to convert a story feels more present. As this comic book series idea evolves throughout the entirety of the images tested, it continues to be more nonsensical. In this way, I feel like the content here is what I was asking for, and took this as a departure from the style and composition. 
6. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, fish having an underwater tea party with birthday hats, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50

7. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: fish having an underwater tea party with birthday hats, Theodor Seuss Geisel, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50

8. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: fish having an underwater tea party with birthday hats, Alice in Wonderland tea party, Theodor Seuss Geisel, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50

(9-13) From this point, I began developing two compositions - one within the comic book style and the other to be more reflective of a specific imagined scene within this world. At the same time, I began switching out certain words and adjusting their order, in particular, I was curious about what I could actually get the program to input accurately as what it felt was important to include and I was attempting to remove all human forms/likeness. The color and line consistency, I felt were on track for what I was looking for, and the continued element of visible sea life even in imagined forms was very successful throughout this process. 

9. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: fish having an underwater tea party with birthday hats, Alice in Wonderland tea party, Theodor Seuss Geisel, cartoon comic book multiple scenes, film strips, no humans, shark --ar 16:9 --stylize 50

10. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Party scene with birthday hats, Theodor Seuss Geisel, Alice in Wonderland tea party, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50

11. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Party scene with birthday hats, Theodor Seuss Geisel, cartoon comic book multiple scenes, film strips, no people --ar 16:9 --stylize 50

12. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Fish throwing a party with birthday hats, Theodor Seuss Geisel, cartoon comic book multiple scenes, film strips, no humans --ar 16:9 --stylize 50

13. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, fish throwing a party with birthday hats, underwater school of fish, shark, jellyfish, stingray, cartoon comic book multiple scenes, film strips, no humans --ar 16:9 --stylize 50

(14-24) Was a particularly interesting set of images - individually and in the grouping of four variations. The reason why I felt that this set of images really showed a compelling forward motion or was of special interest was because despite not having the comic book strip lines, the program did still visually divide these compositions. In this way, I felt like the software was storing a sort of muscle memory of the information that was previously input from other queries. We can see a lot of movement but also the water continually broken or separated by this white or light section. Around this point, queries also began chaining, still alternating the order of words and slightly modifying the composition. I was hoping to get my images to display an underwater tea party of sorts within this simplified style. The term tea party took the queries in a weird inventive dinner party compositional direction. In that way, I don’t know that I would say it was the most successful render of the query. Style-wise, Midjourney also began pulling other cartoon characters that were not requested or taken from either of the attached images. A few of them look significantly more like Popeye characters for example. Throughout the progression, there is also a minimizing of the use of yellow. 

14. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats --ar 16:9 --stylize 50 --style raw - Image #1

15. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats --ar 16:9 --stylize 50 --style raw - Image #2

16. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats --ar 16:9 --stylize 50 --style raw - Image #3

17. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats --ar 16:9 --stylize 50 --style raw - Image #4

18. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats --ar 16:9 --stylize 50 --style raw - Variations (Strong)

19. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats --ar 16:9 --stylize 50 --style raw - Variations (Strong)

20. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats --ar 16:9 --stylize 50 --style raw - Variations (Strong)

21. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats --ar 16:9 --stylize 50 --style raw - Variations (Strong)

22. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats --ar 16:9 --stylize 50 --style raw - Variations (Strong)

23. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats --ar 16:9 --stylize 50 --style raw - Image #1

24. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel cartoon comic book multiple scenes, underwater tea party with sea of fish and party hats, no humans --ar 16:9 --stylize 50 --style raw - Remix

(25 - 32) This was by far the most successful step during this process before the final (33-38 compositions). The reason for this was that the use of color, compositions, and snakes felt much more in line with what I was looking for despite it not including a tea party. At some point in the queries, the program rejected my use of keywords such as “chaos” and “weird” along with several other images. Additionally, this was the point where I felt finally like I had two very tangible compositions including a comic book-style set of images that progressed in some way. 

25. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50

26. https://s.mj.run/fgUOrk1Ilq4 :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, fish throwing a party with birthday hats underwater, no humans, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips, no humans --ar 16:9 --stylize 50

27. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Image #2

28. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Image #4

29. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Variations (Strong)

30. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Variations (Strong)

31. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Image #1

32. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Image #3

(33-38) At the end of this visual exploration, I did have at least one of my final two versions including a comic book strip. In some ways, I felt that Midjourney generally understood what I was asking but missed some crucial aspects that I was looking to have included. This of course makes sense for the level of complexity and detail I was applying to the queries to make this happen. The comic book strips look fine at an initial glance, however, after closer inspection, we can see where there are blatant visual gaps. This is most apparent in the departure of the subjects rendered - they are hardly distinguishable as subjects and at times glitch out. They lack sense. The scenes don’t particularly tell a story, narrative, or scene progression either. With this in mind, they do capture the general style, line weight, and color I was looking for. It feels like this is an inventive Dr.Suess scene. The final frames lack many very specific details of what I asked at various points, I don’t think either represents a tea party scene but overall I am excited about the direction I was able to achieve. The final single frame would be the most compelling and visually accurate version of what I hoped would happen.

33. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Zoom Out

34. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Image #3

35. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Zoom Out

36. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Image #1

37. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Variations (Strong)

38. https://s.mj.run/iMtKAPVuZXM :: https://s.mj.run/T-iOLQw8KP0 :: Theodor Seuss Geisel, no humans, fish throwing a party with birthday hats, underwater school of fish, jellyfish, cartoon comic book multiple scenes, film strips --ar 16:9 --stylize 50 --style raw - Image #1

ASSIGNMENT
For this assignment we are continuing with MidJourney but introduce img2img, a technique by which one or more images are used to generate a new image. Additionally, there is also the added function of a text prompt.

Recommendations for the assignment - Try combining 2,3,4, etc. images as your image prompts. Post a minimum of 2 images per multiple image prompts studies. For instance, you might want to test which combination is of interest to you, 2,3,4,etc. If any of these seem to give interesting results, then stay with the combination and test diferent ways to combine the source images, as the order of which image is listed first in the prompt may determine what the results will be like.

What may be other conceptual questions in using multiple images? Try to go beyond conventional results with the intent to test what may be limits of how MidJourney works. Will MidJourney produce total noise? Will it organize the image space based on visual elements from each of the image prompts? ...Will it try to impose a style, or aesthetic? In the end, your goal is to arrive at results that define your aesthetic interests, not the original images', nor the MidJourney aesthetic.

Series 1: Prompt: https://s.mj.run/274y95Xf-58 https://s.mj.run/hS9wdd5amNA https://s.mj.run/yWqBKs9stmM collage, maximalism, dreamscape, fuzzy camera, surrealism --ar 16:9 --style raw --s 250
My images are all reminiscent of early 2000s digitalism "frutiger metro/aero" aesthetics. This first prompt was an attempt to brainstorm directions to subvert the aesthetic somehow. I was pretty happy with the images. I liked that they blended realism and surrealism in a cartoon-y style. "Collage" was initially intended to generate lots of images to branch off of, but I liked the visually interesting layout so I kept it in future prompts.

Series 2: Prompt: https://s.mj.run/274y95Xf-58 https://s.mj.run/hS9wdd5amNA https://s.mj.run/yWqBKs9stmM collage, maximalism, surrealism, dreamscape, fuzzy camera, mood lighting --ar 16:9 --chaos 70 --style raw --s 250
I kept the same images and some of the same keywords. For this prompt, I wanted to see if I could get a clear distinction between realistic and digitally-made elements in my image. It wasn't that successful -- the generated image looks like a regular old postcard. I can definitely see the collage element, though. Some of the randomness and overlapping of the element positions probably came from my source images, which have lots of overlays which contribute to their busy look.

Series 3: Prompt: https://s.mj.run/274y95Xf-58 https://s.mj.run/hS9wdd5amNA https://s.mj.run/yWqBKs9stmM collage, portal from 2D digital world into reality, maximalism, surrealism, dreamscape, eyes --ar 16:9 --no people --style raw --s 250
This prompt was more literal with the phrase "portal from 2D digital world into reality". I wanted to see a clear difference in style between the digital world and the real world. This was one of my more successful prompts. I didn't even have to add source images for the realistic nature scenes shown in the background. I also liked the eye motif and added it to my prompt to represent "seeing" beyond the screen.

Series 4: Prompt: https://s.mj.run/274y95Xf-58 https://s.mj.run/hS9wdd5amNA https://s.mj.run/yWqBKs9stmM collage, maximalism, dreamscape, fuzzy camera, surrealism, randomness --ar 16:9 --style raw --s 250
I backtracked for this series. I went back to the image with the cat from series 1 because I liked the random placements of the elements. The way the cat looks like it's floating creates the illusion that it's at a different depth within the image despite the image being 2D. I was happy with the results of this prompt, specifically, the realism of the cat in contrast to its digital landscape. I wanted to explore this juxtaposition further in my next prompt.

Series 5: Prompt: https://s.mj.run/274y95Xf-58 https://s.mj.run/hS9wdd5amNA https://s.mj.run/yWqBKs9stmM collage, mixed media, maximalism, dreamscape, fuzzy camera, realism, digitalism, surrealism, randomness --ar 16:9 --style raw --s 250
To tease out the juxtaposition of realism against digitalism, I added "mixed media" into my prompt. This produced an overlay of digital images on realistic nature settings. I like the idea of imposing digital representations of real-world objects in the real world. The bird, for example, is obviously digitally rendered but is placed among organic rocky textures and flowers, where real birds would be found.

Series 1
https://s.mj.run/-nb30pxQbvc https://s.mj.run/Wc6aBov-_uU https://s.mj.run/INgJzpchCwY --weird 500 --chaos 50 --ar 16:9 --style raw --s 250

With my first series, I noticed that MidJourney would attempt to impose a style as long as I added text prompts after the image prompts. In keeping my results in line with the photos I chose, I decided not to use any extra text prompts, and stuck to scaling weird and chaos parameters, as well as prompt weights. With all of my pictures in these series, I wanted to go for an early 2010s aesthetic, coinciding with the rise of social media. This first series looks like webcam screenshots but with dark, surreal elements. I really liked the second image of the four variations, I felt that they all complemented one another. This series is great but I felt the photos weren't telling me enough in terms of motifs.

Series 2
https://s.mj.run/DNyCKiL0CVE https://s.mj.run/ENKlgjIb2rI https://s.mj.run/S1xp2kfm4FY --weird 250 --chaos 20 --ar 16:9 --style raw --s 250

In this series, I wanted to focus on negative space and digital glitching. The lack of facial features was an interesting element of these images, and is an element that is carried throughout most of my work. I think the lack of faces connects viewers to the images as they work their brains to create a face to the faceless, heightening immersion. I also feel like the lack of facial features helps signal to viewers that the image is not real, and therefore reduces the uncanny valley response that MidJourney can create.

https://media.discordapp.net/attachment ... eight=1020

Series 3
https://s.mj.run/DB1Q1ksfD_k https://s.mj.run/M9w0JxLPywY https://s.mj.run/duWbElfR9s0 --ar 16:9 --style raw --s 250

I wanted to experiment with bright flash on a foreground subject. I ended up getting the effect I wanted, but the clothes were difficult to manipulate. I tried to add different weights to each image prompt, and I also tried to add text prompts, but I was unable to get my vision to work. The following images were still very interesting, with one notable element being the complete paleness of the subject. From pale skin to pale clothes, It's almost like these could be photographs of ghosts.

Series 4
https://s.mj.run/1MzEAFhy6iE https://s.mj.run/A2B54her-qk https://s.mj.run/qSXUymhWcNo https://s.mj.run/4vav-TXUFI8 https://s.mj.run/S1xp2kfm4FY https://s.mj.run/DNyCKiL0CVE https://s.mj.run/54UZ4k0DLH4 https://s.mj.run/M9w0JxLPywY --ar 16:9 --style raw --s 250

I had a blast with this prompt as I combined 8 photos and tested different weights and parameters. I was able to play with digital glitching, bright flash, negative space, facelessness and texture, all in one. To me, these images evoke a dreamlike introspection, as if they're random memories of the past, trying their best to be recalled properly.

Series 5
https://s.mj.run/e7RBm4eH-d8 https://s.mj.run/oXOMIpUxGdA https://s.mj.run/cGYH_PAX6kY https://s.mj.run/FdEsj6xHyUw https://s.mj.run/Fg8N6oGmBzA --ar 16:9 --style raw --s 250

This last series I felt was very beautiful, as every image has this foggy, soft look to it. The flower motif is very interesting, as it can signify the existence of an event or ceremony. The flowers alongside the use of negative space created by glitches, create a contrast of nature and technology, which is something I really enjoy experimenting with. Perhaps these images are actually signifying a marriage of the two, or maybe they're signifying the departure of their relationship.

https://media.discordapp.net/attachment ... eight=1020

https://media.discordapp.net/attachment ... eight=1020

Series 1: Prompt: "https://s.mj.run/le9v1--wsqQ , https://s.mj.run/R6pX3gBNl1U --style raw --s 250" The first series was to see how the bot would combine two images without any further user input other than the default raw and s 250 filters. It's interesting to see how in this series, the bot doesn't depict the people as people, but rather converts them into MOSTLY food products, almost as if the images I chose to upload resemble a sort of commercial. I chose to use a street view of an American city, as well as add in a similar neon art piece. There is a heavy bias for orange colors is present here, as there is almost no dark color added from the outrun image, but the sunset was made more intense. It's possible the reason it turned the images into these sort of product promotions is due to how often high end luxury items, such as Jewelry, or overpriced sweets, are advertised with similar camera angles or background scenery.

Series 2: Prompt: "https://s.mj.run/le9v1--wsqQ , https://s.mj.run/R6pX3gBNl1U , https://s.mj.run/E8OT3YpItSI --style raw --s 250" I decided to redo the first two images, and add in an image of head shaped busts. The blue colors became a lot more apparent, and the scene shifted greatly to a blue bias. Once again we have several high end seeming, luxurious items being displayed over a city view, with an outlier that depicts a far different scene. It appears as though the bot did not recognize the shapes as human heads, as it seems all of the images only lifted color from the 3rd image. However, the major outlier is curious, as it preserves the camera angles that the first two images have, and adds in foliage. The light amount of green introduced combined with what I can imagine being the outrun image's dark cool colored ground, with the blue from the heads, made the bot think of nature and a forest with a river running through it. Could the bot associate green and blue combinations from what it interprets as "abstract forms" (the busts it could not comprehend as human)?

Series 3: Prompt: https://s.mj.run/Vzzd2t8XRLg , https://s.mj.run/wAXvN30npBA , https://s.mj.run/2Lk5njr3Sog --style raw --s 250 Without any text input, the bot once again made advertisements. I am not sure if the placement of the green rings in one of the images has anything to do with why it made more product ads, but this time it didn't use the city as a background. It barely used the orange forest's original concept at all at a glance. The forest was turned into the orange copper in the items, as well as produced the leaves in the first image, but the blue water was more prevalent, and it begs the question if the bot is much more biased toward blue than orange. I'm unhappy with the fact it produced more advertisements, and not a sort of forest with green and blue hues on top of the orange. It appears the bot is biased toward making advertisements, as that may be what the developers intend to market it as. A cheap way for companies to produce advertisements.

Series 4: Prompt: https://s.mj.run/qXNr-ni9uG0 , https://s.mj.run/2Lk5njr3Sog --style raw --s 250 For my final prompt, I chose to refer to the original variation in series 2, and combined it with the forest. It once more, produced an advertisement, but at least it produced more varied results. 3 of the images were stylized pieces of art, but still had that sort of front and center feel that the advertisement images did, at least the first one of this series does. The 2nd and 4th feel much more like an actual blend of the 2 images provided, without the out of place "products." The first image is ambiguous enough to seem as just a piece of art, not an advertisement, but it could be candy made for a Halloween seasonal period (a possibility given the association of orange leaves, fall, and Halloween).

I decided to redo some of my prompts, twice each, three times my third series, and see how much they change from an advertisement seeming image, to other scenes. With more runs, it appears the bot made different decisions, and was actually able to recognize the human face in series 2 after more runs, but it kept other habits like turning the green in the busts to foliage. The series 1 redo's remained largely as advertisements, only deviating to emphasize the sunset and a dog in their own versions. Series 3 I added specific prompts to each, and the bot only maintained front and center object advertisements when I used just one word for the filter, but on the one I used two words, the scenes were much more deviated from the other iterations of that series. The bot if left uncontrolled by text input, appears to have a significant bias to focusing on a single object in a scene, and more often than not it does so in a way that resembles an advertisement.
Quick Re-Prompts:
Series 1: https://s.mj.run/le9v1--wsqQ ::2 , https://s.mj.run/R6pX3gBNl1U night time--style raw --s 250 --style raw https://s.mj.run/le9v1--wsqQ ::0.5, https://s.mj.run/R6pX3gBNl1U night time--style raw --s 250 --style raw https://s.mj.run/le9v1--wsqQ ::2 , https://s.mj.run/R6pX3gBNl1U outrun underwater--style raw --s 250 --style raw https://s.mj.run/le9v1--wsqQ ::0.5 , https://s.mj.run/R6pX3gBNl1U outrun underwater--style raw --s 250 --style raw



Series 2: Series 3:
original prompt + uncanny original prompt + uncanny horror original prompt + horror

Prompt Reference Images







Prompts
https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec overlay images --ar 16:9 --style raw --s 250

https://s.mj.run/Q4DU84ronpE https://s.mj.run/4RDFPrjRpoY metal fruit, orange --ar 16:9 --style raw --s 250
I try the orange reference image to refer to it directly in the text portion

https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec grafiti on photo of joe biden inauguration, ink, graphic --ar 16:9 --style raw --s 250

https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec graffiti on top photo of joe biden inauguration, ink, graphic --ar 16:9 --style raw --s 250

https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec graffiti top overlay, joe biden inauguration bottom overlay --ar 16:9 --style raw --s 250

https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec graffiti top overlay, joe biden inauguration bottom overlay --ar 16:9 --style raw --s 250
I introduce first ref image of overlay

https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec --ar 16:9 --chaos 100 --style raw --s 250
I introduce chaos and remove the text prompt



Images

https://s.mj.run/Q4DU84ronpE https://s.mj.run/4RDFPrjRpoY metal fruit, orange --ar 16:9 --style raw --s 250
I try the orange reference image to refer to it directly in the text portion



https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec overlay images --ar 16:9 --style raw --s 250



https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec --ar 16:9 --style raw --s 250 - @MAT 255 (fast)



https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec --ar 16:9 --chaos 100 --style raw --s 250 - @MAT 255 (fast)



https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec --ar 16:9 --chaos 100 --style raw --s 250



https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec --ar 16:9 --chaos 100 --style raw --s 250 - Image #1 @MAT 255



https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec --ar 16:9 --chaos 100 --style raw --s 250 - Image #1 @MAT 255



https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec --ar 16:9 --chaos 100 --style raw --s 250 - Image #2 @MAT 255



https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec --ar 16:9 --chaos 100 --style raw --s 250 - Image #3 @MAT 255



https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec --ar 16:9 --chaos 100 --style raw --s 250 - Image #4 @MAT 255



https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec --ar 16:9 --chaos 100 --style raw --s 250 - Variations (Strong) by @MAT 255 (fast)



https://s.mj.run/cGxyj95fkUE https://s.mj.run/4RDFPrjRpoY https://s.mj.run/N4HZWRKyAec close portrait --ar 16:9 --chaos 100 --style raw --s 250 - Image #2 @MAT 255



https://s.mj.run/4RDFPrjRpoY https://s.mj.run/cGxyj95fkUE https://s.mj.run/N4HZWRKyAec --ar 16:9 --chaos 100 --style raw --s 250 - Image #1 @MAT 255



https://s.mj.run/cGxyj95fkUE https://s.mj.run/N4HZWRKyAec https://s.mj.run/4RDFPrjRpoY --ar 16:9 --chaos 100 --style raw --s 250 - Image #2 @MAT 255



https://s.mj.run/cGxyj95fkUE https://s.mj.run/N4HZWRKyAec https://s.mj.run/4RDFPrjRpoY joe biden inaugeration, grafiti overlay --ar 16:9 --iw 2 --chaos 100 --style raw --s 250 - Remix by @MAT 255 (fast)


Reflection
This was by far the happiest I've been with getting results from Midjourney. My main interest going into experimentation was to create a punk aesthetic, specifically aiming to create images that look like they've been graphically hand-edited in a layered fashion. To accomplish this my goal was to give it writing to overlay on an image or scene. As we've previously noted, Midjourney greatly struggles with text and typography. To address this I attempted to use graffiti tags as the text and included one in the reference prompt images. This ultimately proved very successful with how Midjourney would interpret and produce typography.

After it failed to understand the concepts of overlay and layering, I took a step back and tested it with a new prompt reference image (the orange) to better gauge how it interprets the images together and in relation back to the text portion of the prompt. With those results, I proceeded to find an image that visually showed an example of overlay to include as a reference prompt image. This had an impact on taking the weight away from some of the other images, but what had a greater impact was introducing the chaos parameter. Midjouney began to stray away from its built-in style and began to offer some results getting closer to what I was looking for in aesthetics. What I tested next was removing the text portion of the prompt entirely. What I would've expected with the lack of control was for Midjourney to impose its style much more heavily, but instead with the images, chaos, and repeated trials I began to get results that I was looking for. I kept pushing this study and kept getting more and more of what I was looking for across a wide range.