Media Arts and Technology

Project 1: MidJourney, due: Thursday, October 5, 2023

Please post 4 to 6 images here with their text query. Following that, do an analysis of the results. For each image discuss the following:

1. Does the image meet your expectations?
2. To what degree does your text query influence the generated image?
3. What is the style of the image, and why do you think it has produced that?
4. Any thoughts about how the visual elements in the image are organized
5. How would you change the query? to achieve what difference?
6. Any other comments?
7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result

--
https://discord.com/
https://docs.midjourney.com/docs/variations
https://github.com/willwulfken/MidJourn ... -Reference

Reference article:
https://arstechnica.com/information-tec ... ine-again/

Prompt: "The inside of the brain visualized, web structures, cyberpunk, recursive --ar 16:9 --weird 500 --s 50"

1. Does the image meet your expectations?
Yes. I was expecting an abstraction of thought processes. I can see how this image depicts a mechanical/chemical representation of human thought. Its symmetry, precise lines and angles, and apparent bursts of light remind me of machine blueprints and chemical reactions. I understand this interpretation as the brain can be conceptualized as a chemical machine.
2. To what degree does your text query influence the generated image?
The image's style and composition are the main factors influenced by my text query. I believe the image's flashy colors and industrial edge can be attributed to the word 'cyberpunk' in the prompt. I can also see how some aspects of the image are self-similar, likely due to the use of 'recursive'.
3. What is the style of the image, and why do you think it has produced that?
The image is maximalist, intricate, and precise. I think these are a result of my prompt's keywords. 'Web structures' is responsible for the thin spider-silk lines and precision, 'cyberpunk' is responsible for the image's neon colors and techno-industrial shapes, and 'recursive' is responsible for the image's symmetry and intricacy.
4. Any thoughts about how the visual elements in the image are organized?
I like that the elements seem to burst from a center origin point. This reminds me of the origin of ideas as the burst can be seen to represent the formation of a thought.
5. How would you change the query? to achieve what difference?
I would change the query by including more organic-sounding words. I would like to see thought processes represented in looser, more natural styles.
6. Any other comments?
The image resembles the particle accelerators shown in sci-fi movies. It's interesting to imagine that humans may have modeled creations like these after their own mind structures. Then again, it's more likely that it's the other way around: the AI's representation of the human brain is based entirely on what the human brain has produced. The only way we can conceptualize our mind is with the mind itself, so there's no way to prove or disprove any representation.
7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
4 - the image is visually captivating and makes sense. It reminds me of the floaters I see when I rub my eyes. Surely there must be merit to how accurate this image is to biology.
Prompt: "The inside of the brain visualized, web structures, cyberpunk, recursive --ar 16:9 --weird 500"
1. Yes, the image meets my expectations. I expected an image with lots of ambiguity and complexity to represent the ambiguity of thoughts.
2. My text query influenced this image to a high degree. I can see the web structures and self-similarity in its symmetry. I also would credit 'cyberpunk' for the neon colors and graphics.
3. The style of the image is vaguely medical. It seems to resemble a womb. I think this is because the brain is associated with biology and the origins of thoughts can be thought of as 'births' in themselves.
4. Again, the symmetry and burst from a single origin point in the middle reminds me of a womb. There's also some randomness on either side of the origin point, which makes sense to me as brains will evolve based on environmental factors.
5. I would change the query by rewording it to be less physical. For example, instead of "The inside of the brain", I would write: "The brain as a ___" with the hopes of achieving a less literal interpretation.
6. N/A
7. 4 - I like the complexity of this image. It's visually stimulating and reads like a puzzle. Chaos and randomness is masked by general symmetry and order.
Prompt: "The birth of a thought, webbed structures, biological, photorealistic, medical --ar 16:9 --weird 100 --s 50"
1. The image kind of meets my expectations. I didn't expect it to be so literal.
2. My use of 'biological', 'webbed structures', and 'medical' definitely contributed to the object's structure. I could see it being a graphic in a medical journal.
3. The image reminds me of the webbed skirts found on some species of mushrooms. I think this was produced because I used 'biological' and 'photorealistic'. The AI was probably referencing nature images. I wonder if it would have drawn from mushrooms had I not used the word 'webbed structures' since mushrooms are often associated with sentience and awareness for the psychedelic properties that some have.
4. I like that the object is centered in the image with a minimal background. It places all of the focus on the object.
5. I would remove 'webbed structures' from the query to see if the AI also associates thoughts with webs, as I do.
6. N/A
7. 3 - the image is quite literal.
Prompt: "The creation of sentience, lithography, blueprint, mechanical --ar 16:9 --chaos 80"
1. No, the image doesn't meet my expectations. I was expecting a different representation of sentience. I felt like the use of a human was too literal of a representation for sentience.
2. The words 'blueprint' and 'mechanical' contributed to the steampunk-style of the image. 'Mechanical' is probably why the image's background resembles a CPU. I'll lean into this steampunk aesthetic in my future prompts.
3. The style of the image seems almost biblical. There appears to be a halo around the woman and she is connected to a circular object that looks astrological. Possibly because I used the word 'creation', which is commonly associated with biblical themes.
4. I like the image's background. It adds a technical nuance to something that seems so spiritual at first glance. I'm a little thrown off by the woman's nakedness. I wonder if this is a manifestation of the control exerted over women, who are sometimes believed to be the 'creation' or byproduct of men.
5. I would add more specific terms to my query to make it less likely to generate anthropomorphic images. I wanted a perspective on sentience itself, not the human it is attached to.
6. N/A
7. 3 - the image is not abstract enough for my taste.
Prompt: "Blueprint of a human, stencil, black and white, medical, steampunk --ar 16:9"
1. Yes, the image meets my expectations. I expected a blend of anatomically accurate and mechanical components.
2. I think the words 'stencil' and 'steampunk' caused the pencil sketched blueprint style.
3. The style of the image is as expected. It resembles a blueprint of a portrait or medical journal graphic.
4. So far, most of my images have been centered and vaguely symmetrical, like this one. I would like to see if I can create less conventional compositions in the future.
5. I would include different angles in my query to produce less conventional compositions.
6. N/A
7. 3.5 - the image is visually interesting, but the concept is not as abstract as I would like it to be. Also, from far away, it looks just like any other anatomy sketch.

Prompt: man on fire on stage, crowd cheering, upper vantage view --ar 16:9 --s 50
The image meets my expectations to some extent while also subverting them. My main goal with this prompt was to have a figure of a man on stage who is on fire. Instead, despite rewording and rearranging the prompt I was often given similar-looking results of a man on stage with some lights and fire in the background. However, this result differed from the rest creating an interesting atmospheric effect. My text query most definitely had an influence on the generated image, but not to the extent where I was able to get it to the specifications of what I had drawn up before. The style of the image is realism, though a number of the other images produced were of a more illustrative style. My appeal to this image primarily has to do with the intertwining of elements between the people, actions, and events. I would like to change the query to the original prompt I was attempting to try and imbue more of a storytelling element. However, it should be said the arrangement of elements in this produced image also offers some narrative, which is why I selected it. Overall, I would rate it a 3.
Prompt: gender neutral person crying, smoke instead of tears --ar 16:9 --s 50
This image did not meet my expectations. This is where I am starting to feel some of the difficulty with the prompts that occur when trying to mesh more than one element with another. The style of the image is depicted as realistic likely because of the emotions depicted in the prompt and the element of smoke. I do think that the AI did do more of a successful job in depicting a gender-neutral person than what I have seen in the past. I would like to change this so the AI loses the center composition it gravitates towards and hopefully make the elements meld with each other. I do enjoy the representation of the colors with the depicted emotion. I would rate this a 2.
Prompt: kid running for class president, hope style poster from obama campaign --ar 9:16 --s 50
This did not meet my expectations at all. There was some amount of subversion of my expectations to the benefit of the outcome like the AI interpreting the query of “running” literally. The style seen is depicted as realistic despite the efforts of the query. Something that became clear to me was similar to the cultural code imbued with prompts in general where they typically assume the cultural dominant image of the white man. This particular query with the word “obama” resulted consistently in images of people with darker skin complexions. The norm remained with male representation. I found it interesting that the AI tried to return text in the image. This particular one with the element of text was the best result because the composition hid most of the text’s imperfections. I would rate this a 3 or 4 particularly because of the interesting takeaway of the cultural influence and how it can be navigated in ways other than direct negations to the dominant cultural norm in the prompt
Prompt: extreme close up of teddy bear head, right eye in center --ar 16:9 --s 50
This image did not meet my expectations. The prompt directly influenced the resulting composition, which the result(s) consistently replicated. However, the prompt did not influence the style. The aim of this prompt was for it to be a teddy bear eye—so this wasn’t exactly the expected response. I found myself drawn to this image because the elements of the hair, texture, and light are of interest to me as an artist and the work I do. I would like to change the query and result to be more at the intent of the original query of a teddy bear eye as I believe it would stray away from the more cliche image of the eye. I would rate this a 2.

Image 1: Please post 4 to 6 images here with their text query. Following that, analyze the results. For each image discuss the following:

1. Does the image meet your expectations?
No, I do not feel like this met my expectations. With some of the key works like drawing for example, I was expecting a looser style with more lines, rather these are very realistically rendered - which I feel was not my intention. I was expecting a greater inclusion of “tourmaline” even if it wasn’t included in an eye. I imagined that the shark would have a greenish hue or that only the eye would be that green color.

2. To what degree does your text query influence the generated image?
As far as visual elements in an organization, the great white shark and black and white aspect of my query captured the strongest.
I feel the image generated did make sense with the text I had input, however, it was not the direction that I was envisioning personally.

3. What is the style of the image, and why do you think it has produced that?
I think based on nature imagery and how sharks are typically rendered through photography, documentaries, tattoos, or art in general - I see how the AI-derived this interpretation from my query. I was hoping it would be more cartoon-like and significantly more whimsical.

4. Any thoughts about how the visual elements in the image are organized?
The visual elements do seem to be organized to be in line with the query I attempted. It appears that Midjourney did have a gap in some of the things that were included or not. The visual element of including anything related to a tourmaline (in color or physicality) did stump me a bit.

5. How would you change the query? to achieve what difference?
I would change this query to attempt to better combat the influence or power of AI stylization. I feel like that is very apparent and I would like something more imaginative and cartoon-like.

6. Any other comments?
Based on this query, I did make significant changes or jumps to the query to try other styles but was still not able to achieve what I wanted exactly - which resulted in further straying or abandoning this idea in part unfortunately.

7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
2.5/5

Image 2: Please post 4 to 6 images here with their text query. Following that, analyze the results. For each image discuss the following:

1. Does the image meet your expectations?
This image still did not meet my expectations exactly. I wanted an image that was imaginative, playful, and maybe rendered even a bit flatter. This doesn’t necessarily feel like the same sense that Dr.Suess’ drawings and/or illustrations typically evoke. They seem a little darker and in some ways too adult for the concept requested.

2. To what degree does your text query influence the generated image?
The query does seem to be a bit of a miss. Only one out of the four images depicts a proper shark despite it being listed as the first word - Midjourney instead began to generate the term “shark” mores as lizard type creates out of a body of water. These seem like weird adult swim cartoons - which was not the direction I was hoping for.

3. What is the style of the image, and why do you think it has produced that?
The style of the images varies significantly but overall are rendered less realistically and more cartoon-like or illustrative. One of the keywords I entered in this query was to include a Dr. Suess style - this mark was missed significantly, however. The images produced missed the stylization process of creating 2-D flat illustrations with minimal color and dark line work. All of these have too much shading and depth for what was being asked. Again, I feel that the software had too much of a stylization of the outcome of the query.

4. Any thoughts about how the visual elements in the image are organized?
Despite the phrase “drinking a shrimp cocktail” being listed last - it seems that this was the element that was actually the most successfully generated of the query.

5. How would you change the query? to achieve what difference?
Based on this query, I would try to change it to be much looser, cartoon-like, and more minimal in design. I am looking for a flatter image with thicker lines and way less detail. In the next query, I am going to try other styles from different artists or find other ways to minimize this level of detail. The shrimp cocktail will also be omitted for the next revision. Even though I think this was the strongest part of the query, I want the visualization to be on the shark - not the drink.

6. Any other comments?

7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
3/5


Image 3: Please post 4 to 6 images here with their text query. Following that, analyze the results. For each image discuss the following:

1. Does the image meet your expectations?
To a certain degree - yes. There are still some elements that I don’t feel generated the way that I would’ve expected or hoped in an outcome.

2. To what degree does your text query influence the generated image?
The further into generated images that we are getting with this assignment, the more I question if the order of the words matters that much to the image creation. I feel like Midjourney is pulling the elements that it understands the most or has the largest database of to merge first and foremost. This set of four images, in some ways, captures my query but two of them don’t properly depict the tree or the placement together with the shark.

3. What is the style of the image, and why do you think it has produced that?
This style is a bit more surrealistic - based on the text I would’ve imagined this to be the case. In the previous version, I chose a style that was more indicative of illustration works, however, that was not the case for this query. The author, Jonny Sun is a poet, writer, and illustrator - I thought based on that information there was a chance that Midjourney could capture elements of the realistic nature of his written works or his illustrations. Overall this still feels extremely stylized, I can see the definitive influence of Midjourney’s style on these images and that is something I am currently having a hard time combating.

4. Any thoughts about how the visual elements in the image are organized?
Overall the visual elements incorporated are the most successful here of all the queries thus far. I am a bit perplexed about the inclusion of the tree above the water, this is something that can be further pushed. The term chaos was maybe skipped or evaded a bit in the final imagery.

5. How would you change the query? to achieve what difference?
In future queries, I would try to create something that omits the sky with clouds and possibly plunges the visualization further into chaos. This was not as chaotic or futuristic as I would’ve liked to see. This feels too simplistic for what I had envisioned.

6. Any other comments?
The trees made for an interesting composition but still missed the mark. The upper right-hand image is probably most what I had hoped for but still doesn’t quite hit what I was attempting to create.

7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
4/5


Image 4: Please post 4 to 6 images here with their text query. Following that, analyze the results. For each image discuss the following:

1. Does the image meet your expectations?
2. To what degree does your text query influence the generated image?
3. What is the style of the image, and why do you think it has produced that?
4. Any thoughts about how the visual elements in the image are organized
5. How would you change the query? to achieve what difference?
6. Any other comments?
7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result


Image 5: Please post 4 to 6 images here with their text query. Following that, analyze the results. For each image discuss the following:

1. Does the image meet your expectations?
Yes and no.
2. To what degree does your text query influence the generated image?
I think the order of the query is not necessarily so important - rather this software is pulling as much information from a database as possible and if there isn’t enough data to pull from I can see how that can’t properly influence the final image. So far the text order has not made a large enough difference for me to see throughout multiple iterations.

3. What is the style of the image, and why do you think it has produced that?
The style is not anywhere close to what I was picturing, if anything some of these seem a bit scary almost for trying to capture something “cutesy”.

4. Any thoughts about how the visual elements in the image are organized?
Overall this image has had the best organization of elements so far. The only miss was the style and lack of Dr. Suess-like elements, imagination, or style.


5. How would you change the query? to achieve what difference?
Most of the elements I listed are captured here, I would keep pushing this more stylistically in the future. Of all the queries thus far, this has done the best of holding all the elements listed even if visual the style isn’t as exact as I’d prefer.

6. Any other comments?
7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
3/5


Image 6: Please post 4 to 6 images here with their text query. Following that, analyze the results. For each image discuss the following:

1. Does the image meet your expectations?
Not exactly. This stylistically still missed the mark I was hoping to achieve.

2. To what degree does your text query influence the generated image?
Still highly stylized and visually influenced by the program at hand.

3. What is the style of the image, and why do you think it has produced that?
Although this is a significant step closer to what I was hoping. None of the images produced are rendered within the world or style of Dr. Suess’ literature. In terms of rendering a children’s literature style though - based on my understanding of the book market, the program did do a good job at reproducing several different contemporary styles among illustrators today.

4. Any thoughts about how the visual elements in the image are organized?
Here the shark is most apparent - which is considerably very good. For the rest of the visual elements, I feel as though much of my query is missing. I was hoping for an underwater city with bustling elements or other creatures. Visually in many ways, this missed the mark in terms of style and element composition.

5. How would you change the query? to achieve what difference?
In the future, I will keep pushing the query and breaking down the defaulting style of the program. There is a more simplistic and less detailed style that I believe Midjourney could create. Especially with how much visual information is linked to Dr.Suess’ works online - it does seem surprising that this was not captured better within his signature style.

6. Any other comments?

7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
3/5

Foreword: Below are four variations of the same prompt I used to create my first ever Midjourney images. The only thing that varies in the prompts below, are the aesthetic stylisation requested for them. I then went ahead and chose 4 images in total (one from each sets of fours) that follows the prompt to its closest approximation. I explain below why I chose those, with a critical explanation of why I chose them over the others as well as in what regards and to what degree the AI image generator on the one hand surprised me by delivering the unexpected, and on the other, did not, in a certain sense, meet my expectations with the prompts I provided.



Image 1: Prompt: A science classroom underwater with fatigued students doing math problems on infinite stair cases, in the style of Japanese lo-fi 64 bit colour, use lighting and shadows --ar 16:9 --c 25 --style raw --s 250 - @MAT 255 (fast)

1. Does the image meet your expectations?
It does something beyond meeting expectations here. On this first ever encounter with an AI Image generator like Midjourney, I realised what it means for the AI to read and synthesise linguistic cues and semiotic signifiers into imagined singnifieds, where the rules of Saussurian semiotics, as it appears in the real world do not hold true. The prompt-based image generation generates goes beyond a structuralist reality to reveal the potentialities of free association between signifiers and signifieds -- one that taps into the loop that feeds the AI engine and lets machinic permutations foster imagined semiotic connections. In short, although the aesthetic parameters in the prompt tries to follow it as closely as possible, the scene description requested by it, given its complexity, has been largely evaded. Even though the rendition of the underwater classroom is fairly accurate, with its use of lo-fi image style and lighting, the infinite nature of the staircase is no where near what was expected by me.

2. To what degree does your text query influence the generated image?
Insofar as the aesthetic stylisation is concerned, the image-style and use of colour requested in the text has influenced the image and rendered the Japanese lo-fi, 64 bit aesthetic quality that I was looking for. I feel the image generated did make sense with the text I had input, however, it was not the direction that I was envisioning personally.

3. What is the style of the image, and why do you think it has produced that?
Based on available examples of Japanese lo-fi art in 64 bit colour, the AI was able to produce the aesthetic quality I was looking for. However, in terms of what I had asked for the scene to be in -- in other words, in terms of its diegetic description -- it has somehow failed to render the infinite nature of the staircases. I was hoping it would be more flat, like an illustration from a book and significantly more whimsical in its rendition of the space.

4. Any thoughts about how the visual elements in the image are organized?
The visual elements, namely the underwater atmosphere of the classroom do seem to be organized to be in line with the prompt I attempted. It appears that Midjourney did have a gap in some of the things that were not included -- namely the infinite stairs. The variation of images, caused by the chaos parameter set to 25 is also quite significantly noticeable and admirable.

5. How would you change the query? to achieve what difference?
I would perhaps change this query to attempt to better combat lack of specificity in the elements of the scene, over the aesthetic style. I would specify the kind of infinite staircase I am aiming for the AI to depict (e.g. the Penrose Stairs).

6. Any other comments?
Based on this query, I did make significant changes or jumps to the query to try other styles but was still not able to achieve what I wanted exactly - which resulted in further straying or abandoning this idea in part unfortunately.

7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
3.5/5



Image 2: Prompt: A science classroom underwater with fatigued students doing math problems on infinite stair cases, in the style of a 3d unity render, Octance render, 3d sculpture, raytrace, 4K, dramatic lighting, focused, detailed --ar 16:9 --c 25 --style raw --s 250 - @MAT 255 (fast)

1. Does the image meet your expectations?
This is one of the rare instances that the image came fairly close to my expectations than the others. The aesthetic parameters in my prompt were followed more or less accurately. The images resemble the style of photo-realistic 3D computer/playstation games -- that sometimes use 3d Unity and Raytrace. Unlike the rest of the examples here, the scenes depicted in these images can be said to be taking place underwater. Although, here again, the scene description requested by it, given its complexity, has been largely evaded. The infinite nature of the staircase is no where near what was expected by me.

2. To what degree does your text query influence the generated image?
Insofar as the aesthetic stylisation is concerned, the 3d unity render, Octance render, 3d sculpture, raytrace, and dramatic lighting effect requested in the text has influenced the image and given it its aesthetic particularity. I feel the image generated did make sense with the text I had input, and the image that was generated are much akin to the virtual games that I based its style on.

3. What is the style of the image, and why do you think it has produced that?
The style of the image is much like that of images from computer games. Based on available examples of effects generated by 3d unity render, Octance render and raytrace, the AI was able to render the 3d sculpture-like aesthetic quality. However, in terms of what I had asked for the scene to be in -- in other words, in terms of its diegetic description -- it has somehow failed to render the infinite nature of the staircases. Image 2 and 4 do not seem to show any sign of staircases at all!

4. Any thoughts about how the visual elements in the image are organized?
This is an example where the visual elements, namely the underwater atmosphere of the classroom seem to be quite well organized to be in line with the prompt I attempted. It appears that Midjourney did have a gap in some of the things that were not included -- namely the infinite stairs. The variation of images, caused by the chaos parameter set to 25 is also quite significantly noticeable and admirable.

5. How would you change the query? to achieve what difference?
I would perhaps change this query to attempt to better combat lack of specificity in the elements of the scene, over the aesthetic style. I would specify the kind of infinite staircase I am aiming for the AI to depict (e.g. the Penrose Stairs).

6. Any other comments?
Based on this query, I did make significant changes or jumps to the query to try other styles but was still not able to achieve what I wanted exactly - which resulted in further straying or abandoning this idea in part unfortunately.

7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
4/5



Image 3: Prompt: A science classroom underwater with fatigued students doing math problems on infinite stair cases, painted by Georges Seurat, use Pointillist brush technique, vibrant colours, epic --ar 16:9 --c 25 --style raw --s 250 - @MAT 255 (fast)

1. Does the image meet your expectations?
As I have mentioned above -- Not quite. Here, even the aesthetic parameters seems to have gone awry. The images do not quite follow Georges Seurat's use of Pointillist brush strokes. The last image in the set of 4 curiously resembles Vincent van Gogh's The Starry Night (June 1889) in its quasi-post-impressionist rendition of the sky. Again, the scene description requested by it, given its complexity, has been largely evaded. The rendition of the underwater classroom is inaccurate, just as the infinite nature of the staircase is no where near what was expected by me.

2. To what degree does your text query influence the generated image?
Insofar as the aesthetic stylisation is concerned, the image-style and use of colour requested in the text has influenced the image and rendered the painting-like aesthetic quality. Though, the style of painting followed was not exactly what I had asked for. I feel the image generated did make sense with the text I had input, however, it was not the direction that I was envisioning personally.

3. What is the style of the image, and why do you think it has produced that?
Based on available examples of vibrantly coloured paintings (not necessarily Pointillist or Post-impressionist), the AI was able to render the painting-like aesthetic quality. However, in terms of what I had asked for the scene to be in -- in other words, in terms of its diegetic description -- it has somehow failed to render the infinite nature of the staircases as well as the the request for it to be underwater. I was hoping it would be follow Seurat's style more accurately and his use of Pointillist technique.

4. Any thoughts about how the visual elements in the image are organized?
The visual elements, namely the underwater atmosphere of the classroom do not seem to be organized to be in line with the prompt I attempted. It appears that Midjourney did have a gap in some of the things that were not included -- namely the infinite stairs. The variation of images, caused by the chaos parameter set to 25 is also quite significantly noticeable and admirable.

5. How would you change the query? to achieve what difference?
I would perhaps change this query to attempt to better combat lack of specificity in the elements of the scene, over the aesthetic style. I would specify the kind of infinite staircase I am aiming for the AI to depict (e.g. the Penrose Stairs).

6. Any other comments?
Based on this query, I did make significant changes or jumps to the query to try other styles but was still not able to achieve what I wanted exactly - which resulted in further straying or abandoning this idea in part unfortunately.

7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
2.5/5



Image 4: Prompt: A science classroom underwater with fatigued students doing math problems on infinite stair cases, in the style of still photographs taken with high contrast Illford HP5 balck and white film, using a 35mm lens, dramatic, Chiaroscuro lighting effect, deep shadows, intense --ar 16:9 --c 25 --style raw --s 250 - @MAT 255 (fast)

1. Does the image meet your expectations?
Not quite. The aesthetic parameters in my prompt were not quite followed. Although, the images are in fact quite high-contrast in nature, thereby trying to emulate the Chiaroscuro lighting effect, with intense, deep shadows, they are not exactly black and white or neutral monochromes, as one would expect from the Illford HP5 balck and white film stock (the lensing is accurate, however). While image 3 has a hint of warm hues in the highlights, image 4 is in subdued colour (almost like in a neo-noir film, like David Fincher's Se7en (1995, USA). Again, the scene description requested by it, given its complexity, has been largely evaded. The rendition of the underwater classroom is inaccurate (image 3, for instance, would fit more the description of a flooded staircase than an underwater classroom). As always, the infinite nature of the staircase is no where near what was expected by me.

2. To what degree does your text query influence the generated image?
Insofar as the aesthetic stylisation is concerned, the use of high-contrast Chiaroscuro lighting effect requested in the text has influenced the image and given it its high-contrast lighting quality. Though, the style of painting followed was not exactly what I had asked for. I feel the image generated did make sense with the text I had input, however, it was not the direction that I was envisioning personally.

3. What is the style of the image, and why do you think it has produced that?
Based on available examples of high-contrast (with or without Chiaroscuro lighting effect), the AI was able to render the photographic aesthetic quality. However, in terms of what I had asked for the scene to be in -- in other words, in terms of its diegetic description -- it has somehow failed to render the infinite nature of the staircases as well as the the request for it to be underwater. I was hoping everything to be at least submerged underwater.

4. Any thoughts about how the visual elements in the image are organized?
The visual elements, namely the underwater atmosphere of the classroom do not seem to be organized to be in line with the prompt I attempted. It appears that Midjourney did have a gap in some of the things that were not included -- namely the infinite stairs. The variation of images, caused by the chaos parameter set to 25 is also quite significantly noticeable and admirable.

5. How would you change the query? to achieve what difference?
I would perhaps change this query to attempt to better combat lack of specificity in the elements of the scene, over the aesthetic style. I would specify the kind of infinite staircase I am aiming for the AI to depict (e.g. the Penrose Stairs). I might also use the word "submerged underwater" instead of just "underwater" in order for it to follow my expectations more closely.

6. Any other comments?
Based on this query, I did make significant changes or jumps to the query to try other styles but was still not able to achieve what I wanted exactly - which resulted in further straying or abandoning this idea in part unfortunately.

7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
2.5/5


Finally, important to note here, is the blue bias in some of images in the first two and the last sets of four [i.e. the fourth image of the last or fourth set]-- something that was not a part of my prompt). Nevertheless, below are 4 images chosen from each sets of four above, that have come more or less close to my expectations, not just in terms of rendering the aesthetic style in the query provided as closely as possible, but more so in retaining most of the visual elements asked for in the scene depicted in the first place. These are:

Image (A)



Image (B)



Image (C)



Image (D)

"grassy hills, blue cloudy sky, bright direct flash digital camera photography, fisheye perspective subject in foreground wearing rick owens balenciaga, motion blur, clipping glitch contortion, lsd dream emulator --chaos 30 --ar 16:9 --weird 100 --style raw --s 250"

1. Does the image meet your expectations?
Yes it does, I wanted to create a scene of someone in a liminal grassy space in the style of old 3D video games and the image met that expectation.
2. To what degree does your text query influence the generated image?
The text query did a good job of influencing this image, however, there are a couple of prompts that failed to affect the image in the way I intended. “Clipping glitch contortion” failed to create any distortion of the subject's limbs or create any glitches. “Bright direct flash digital camera photography” failed to cast a flash on the subject in the way I wanted to.
3. What is the style of the image, and why do you think it has produced that?
I would describe the style of image to be vaporwave/video game-inspired. The prompts for the background I gave it and the text prompt of “lsd dream emulator” worked together to create this liminal space that looks like it was rendered in an old 3d software.
4. Any thoughts about how the visual elements in the image are organized
I really liked the fact that the subject remained in the center of the image, it helps give off the idea that the image is from a video game, and is unnatural. There were some results that placed the subject off-center with natural shading, which both elements in unison create a more realistic looking image, which I did not want. I also love the way the clouds look, creating this ring around the subject, maybe I would have liked to have seen the cloud at the top of the image removed to make the image feel more open.
5. How would you change the query? to achieve what difference?
I would change the query to add “low poly”, “ps1 style graphics” or something to try and add the flash effect I mentioned earlier. I think I can get the image to look like an even older video game from the 90s-00s, and add a flash effect that will make the model look like its glowing.
6. Any other comments?
This was one of my first concepts using Stable Diffusion, really cool to see how Midjourney interprets it.
7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
4



"2010s tumblr bloghouse indie sleaze crystal castles, direct flash blow out photography, fisheye perspective, empty condo interior, subject in foreground wearing rick owens balenciaga maison margiela, volumetric lighting, full body portrait, lsd dream emulator --ar 16:9 --chaos 30 --weird 100 --style raw --s 250"

1. Does the image meet your expectations?
It did, it did not, and also surpassed my expectations. I never told it to create someone holding anything, so it was really cool to see that. It also failed to create the exact background I was hoping for and didn’t cast the flash that I wanted. However, the overall vibe and composition is what I was looking for. I'm not sure if there is anything explicitly "2010s" and "crystal castles" about the image generated, but it is giving off the dark atmosphere that I kind of wanted
2. To what degree does your text query influence the generated image?
I think the text query influenced the image for the most part, there were some things that didn’t get affected with the flash, and the background. Other than that I feel like the “weird” parameter did a good job of making the image abstract, totally on its own. I didn’t ask it to generate any spheres, but the “weird” parameter made a creative spin on the basic composition I wrote.
3. What is the style of the image, and why do you think it has produced that?
I’d describe the image as surrealist photography, many of the prompts I gave it were of real-world photographic styles coupled with fashion. The “lsd dream emulator” as well as the “weird” parameter worked together to give it surrealist details.
4. Any thoughts about how the visual elements in the image are organized
I loved the angle and the fisheye perspective of the image. The walls and columns really show the curved effect that fisheye lenses create. Again I really liked the centered subject, again creating the feeling of deliberate composition,
5. How would you change the query? to achieve what difference?
Again I would try to add words to accomplish a bright direct flash on the subject, so far Midjourney seems to have a lot of trouble with that prompt. I would also maybe add more “weird” to it and maybe change the background to a more modern-looking interior.
6. Any other comments?
This is another concept I’ve tried in Stable Diffusion, the aspect ratio really gives it an interesting look. I’m definitely missing the glitchy, unnatural aspects of version 1.5, but Midjourney so far is blowing me away.
7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
4



"artificial intelligence, osamu sato, lsd dream emulator --ar 16:9 --weird 200 --style raw --s 250"

1. Does the image meet your expectations?
This image surpassed my expectations. I was expecting with my prompt that I would generate some kind of flat 2-dimensional piece of art, however, I did not expect the art to be stylized in the way it was. I was expecting a piece with more geometric shapes but Midjourney generated something that looked more organic.
2. To what degree does your text query influence the generated image?
The text query influenced the image a decent amount when it came to the colors and layout. The organic shapes and figures within the image were a surprise however, as I did not explicitly request for that in my prompt. Also the prompt of “artificial intelligence” failed to reproduce something with any trace of technology, rather it created a nature scene which was really interesting.
3. What is the style of the image, and why do you think it has produced that?
I would describe the style of image to be some sort of digital folk art, or something similar to that of Dutch painter Hieronymous Bosch. I also feel that the color combinations are very hyperpop.
4. Any thoughts about how the visual elements in the image are organized
The organization of elements in this image is very deliberate, with each element having a space of its own. The careful organization of elements reminds me of the organization of computer components. I also feel like all of the little characters and components in the piece represent a co-existing network, communicating with one another akin to that of an AI neural network.
5. How would you change the query? to achieve what difference?
I would add “hieronymus bosch” to my prompt and perhaps change the aspect ratio to 9:16. I think I would like to play more into his style of art, and I think a vertical composition would be more interesting and would make for a fun phone wallpaper.
6. Any other comments?
I really liked this piece, I thought it was really fun to look through and attempt to decipher what was going on and what kind of creatures were roaming around the landscape.
7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
5



"human spiritual connection to the internet, 2010 tumblr flickr bloghouse indie sleaze, intense fisheye perspective, bright direct flash on subject, image taken on low quality digital camera, volumetric lighting, subject in foreground wearing rick owens balenciaga --weird 500 --chaos 50 --ar 16:9 --style raw --s 250"

1. Does the image meet your expectations?
Almost, the visual aesthetic I was envisioning was not met but the image is interesting nonetheless. I was hoping for something more chaotic, and with a darker aesthetic, but this image feels very futuristic to me, which is what I wanted. I also was hoping for some representation of an alien technology that was successfully produced.
2. To what degree does your text query influence the generated image?
Some of the text prompts influenced this image either not at all or in very indirect ways I felt. I do not see fisheye perspective, the image is not pixelated or low quality, nor was the correct lighting effects applied in the image. Rather I see the lighting effects coming from the object the person is interacting with. I’m sure the weird and chaos parameters are the reason for this as Midjourney is failing to apply these prompts in the ways I intended. However, I love the clothes the subject is wearing as a result of the fashion prompt, it really sells the futuristic aspect of my concept. I’m also unsure of how this person is spiritually connected to the internet in this image. The subject is connected to something, though I’m not sure I would describe the object as “the internet”. I would be interested in creating some variations of this image.
3. What is the style of the image, and why do you think it has produced that?
Future surrealism. I feel it produced this mostly due to the internet and fashion prompts, giving the image futuristic and technologically focused elements.
4. Any thoughts about how the visual elements in the image are organized
I like the flat composition of the image, and the simple organization of a single subject and an object. I would have liked to see something more dynamic, creating some kind of motion within the image.
5. How would you change the query? to achieve what difference?
I would be interested in adding motion blur to the query to make the image more dynamic. I would also like to add wires in order to represent a physical connection between a piece of technology (the internet) and a human. Lastly, I would want to mess around with the chaos and weird parameters, lowering them in order for my text prompts to have a chance to actually affect the image.
6. Any other comments?
As much as I want Midjourney to replicate my text prompts, I also want to see to what extent I can stretch the AI into creating something really confusing and surreal. I’ve learned so far that overloading the text prompt is one way to do so, but in the process of making this image, I also learned it’s more interesting to think of specific concepts and to just increase the weird and chaos parameters.
7. On a scale of 5 - from 5 being GREAT to 1 being LOW your rating of the result
3

Prompt: dragon fighting an army of a thousand soldiers --style raw --s 250

1. NO! The images generated lacked a lot of soldiers.
2. The query influences enough to generate an idea, but the specific details of the dragon, the soldiers, the scenery, that is all left up to the bot to decide.
3. The dragon was rendered as a large serpentine creature, similar to dragons from Asian cultures. The overall style of the images is painterly, yet with several small details to make one imagine as not a painting. It's close enough, however, to determine that this is likely due to the age of the stories that involve dragons in this shape, as well as the fact that the soldiers are wielding weapons that resemble medieval ones.
4. The dragon is the largest thing in the image. it's likely this is due to the implied size of dragons in myth, but also possible due to how the dragon word was the front of the query and not the army.
5. I would attempt to make it generate more soldiers, enough to make it seem like there's as much human army as there is a single dragon. The style and shape of the dragon are fine, the lack of humans and perhaps the scenery need change.
6. N/A
7. 3. It was a good render, but not quite what I wanted.



Prompt: an army of humans fighting one dragon --c 50 --style raw --s 250

1. Some images did in fact have what I expected, while others lacked.
2. The query influenced as much as the last one did, except the --c 50 modifier. The chaos likely caused the images to not all have the large amount of soldiers one would imagine at the thought of an army.
3. The first three images followed a very similar painter style, however the last one has a comic style. The comic style was likely produced alongside the lone human facing off against the threatening dragon, likely to call to mind the trope of the lone hero braving tough challenges and taking on terrors and large dangers to the world and their people.
4. The first image shows an interesting battle formation that would likely be used against the dragon. The second image shows what one might assume to be a general, or leader, shouting orders and making sure the army is ready for the fast approaching, already near dragon. The third image is confusing, as there doesn't seem to be much of an army other than the lone armored "hero." It's possible that this shift into a One Man Army led the bot to produce a comic book hero scene for the last image.
5. I would change the query to show what it would do if I asked for a "one man army" specifically. Would it show one person, or several?
6. N/A
7, 4, it was an improvement, and I was content to see the bot bring the "one man army" trope into the mix without having been told to.



Prompt: a speeding car being chased by a donut policeman --c 100 --style raw --s 250

1. None of the images showed the original prompt's idea. Only one even showed a donut and a car. I am not counting the Donut Truck.
2. The query only influenced with the idea of donuts, police, and cars. Never all 3, but sometimes just 2. Policemen and donuts never mixed, and policemen and cars never mixed.
3. Three of four images are cartoony or stylistic, but the last one is more realistic, save for the four finger officer. The style of the donut involved images may be due to how cars and donuts are more of a silly combination, whereas the one with just cars, was more serious in style, but still cartoony. The officer being realistic may just be due to how it only depicts a police officer and nothing else. A more serious scene than a silly idea.
4. The first and third images show actual speeding cars, by blurring surrounding elements in the scene. The other two just place their main object front and center, neither of which depict speeding. It's possible that the officer being on what appears to be a road makes him a traffic cop, thus having a hand in speeding car related issues.
5. I would change the query to remove the chaos modifier and see if it can generate images closer to the original prompt's idea.
6. N/A
7. 2.5/5 due to how it failed to produce exactly as I wanted or anything close, but still provided images interesting enough to study and analyze.



Prompt: a speeding car being chased by a donut policeman --style raw --s 250

1. The images are very close to what I wanted, but they are just off each time. The last one of the 4 showed the closest thing to what I wanted.
2. The query influenced the general idea. Cars, Donuts, Cops. Only one image showed the "chase" part of the query. Overall the query didn't control much other than what objects would be generated, except for scene 4 where it had more influence.
3. The styles are all cartoony, likely due to the idea that a donut chasing a car as a policeman invokes a silly image, thus a cartoony style matches the silly tone of the query.
4. The visual elements are all as expected. The focus is on the donut and the car, or the silly characters next to them. Some images blur the background, while others darken it to maintain focus on the main objects.
5. I would want the query to be changed in a way to produce a realistic image every time. I want to see more of a horrified individual running from a donut at this point.
6. The bot might need a lot more trial and error efforts to get everything about a query right.
7. 3.5/5 this time due to how close it is to the query, but not quite right.