humAI (human & AI) interaction @ reflection

these days, I developed a new habit of browsing Midjourney Showcase every morning to feel inspired with the possibilities of the future.

today while browsing the showcase, i came across an image generated by the Midjourney that immediately drew my attention as a Vietnamese: the beautiful landscape of Halong Bay.

The level of details is amazing, and also as a newly-adopted habit, I started to look at the prompt of the image and surprise at the volume of the prompt, which was:

but as I look closed to the prompt, I knew that something is not right. there are repeated and conflicting keywords. for example, 50mm and 70mm are mostly used for portrait, in order words, it is very hard for you get an “ultra wide landscape” with this feature.

i start to wonder how many of the keywords that the AI can truly understand and make meaning out of it. I came to another friend to ask for advise, ChatGPT 4.0.

Questions:

How many of the below keywords that the Midjourney AI can make meaning and bring to the AI-generated image?

Answer:

The Midjourney AI can interpret and incorporate many of the keywords provided into an AI-generated image, although some may be redundant or overlapping in their effects. Here’s a list of keywords that the AI can make use of:

1. Colonial landscape of the Ha Long Bay
2. Ultra-realistic
3. Cinematic photography
4. Hyper-realism
5. High details
6. 8k resolution
7. Symmetrically
8. Cinematographic
9. Color grading
10. Shot on 50mm lens
11. Ultra wide angle
12. Depth of field
13. Beautifully color-coded
14. Insane detail
15. Intricate details
16. Beautiful color grading
17. Unreal Engine
18. Cinematic
19. Editorial photography
20. Photoshoot
21. 70mm lens shot
22. Tilt blur
23. Shutter speed 1/1000
24. F/22
25. White balance
26. 32k resolution
27. Super resolution
28. Megapixel
29. ProPhoto RGB
30. VR
31. Lonely
32. Good
33. Massive
34. Partial lighting
35. Backlighting
36. Daytime lighting
37. Incandescent
38. Fiber optic
39. Moody lighting
40. Cinema lighting
41. Studio lighting
42. Soft lighting
43. Volumetric
44. Contre-Jour
45. Beautiful lighting
46. Accent lighting
47. Global illumination
48. Ray tracing global illumination
49. Optics
50. Scattering
51. Glow
52. Shadows
53. Harsh
54. Shimmering
55. Ray tracing reflections
56. Lumen reflections
57. Screen space reflections
58. Diffraction grading
59. Chromatic aberration
60. Offset from GB
61. Scan lines
62. Ray tracing
63. Ambient ray tracing
64. Anti-aliasing
65. FKAA
66. TXAA
67. RTX
68. SSAO
69. Shaders
70. OpenGL – Shaders
71. GLSL – Shaders
72. Post-processing
73. Post-production
74. Cel shading
75. Tone mapping
76. CGI
77. VFX
78. SFX
79. Hypermaximalist
80. Elegant
81. Hyperrealistic
82. Super detailed
83. Dynamic pose
84. Ultra-realistic
85. Full-HD
86. High detail high definition
87. 8k resolution
88. AR 16:9
89. V5
    
    Please note that some of the terms may not directly translate into visual elements, but they can help guide the AI to create a specific style, mood, or visual effect in the image.

almost 90 keywords are crazy to think about. but that’s the quantity, how about the quality? i continue to have a conversation with ChatGPT

Question:
How Midjourney may prioritize these above-mentioned keywords?

Answer:

Midjourney AI would prioritize the above-mentioned keywords based on the following factors:

1. Core subject: The primary focus of the image should be on the core subject, which is the colonial landscape of Ha Long Bay in this case.
2. Composition and perspective: Keywords related to composition, such as ultra-wide angle, depth of field, and shot on 50mm/70mm lens, would be prioritized to set up the basic structure and viewpoint of the image.
3. Visual quality: The AI would prioritize keywords that define the desired resolution, detail, and realism of the image, like ultra-realistic, 8k resolution, high detail high definition, and hyper-realism.
4. Lighting and mood: Lighting-related keywords, such as backlighting, moody lighting, cinema lighting, and studio lighting, would be prioritized to establish the overall mood and atmosphere of the image.
5. Color and tone: Keywords related to color grading, beautifully color-coded, and ProPhoto RGB would be considered to give the image its distinct color palette and visual appeal.
6. Technical and visual effects: Keywords related to specific effects or techniques, such as ray tracing, global illumination, shaders, post-processing, and anti-aliasing, would be prioritized to enhance the visual impact and achieve the desired style.
7. Additional attributes: Keywords like lonely, good, and massive would be prioritized to convey the intended emotional tone or scale of the image.
   
   To create an AI-generated image that meets the desired criteria, Midjourney AI would need to find a balance between these keywords and make trade-offs when necessary. For example, it might be challenging to incorporate both ultra-wide angle and 50mm/70mm lens shots in the same image, or to balance a high level of detail with other visual effects. In such cases, the AI would prioritize the keywords that are most essential to achieving the desired outcome, while still trying to incorporate as many of the other keywords as possible.

honestly, that’s a great moment to see that I make a same judgement as the AI! I ask it to continue.

Question: what are some more of the conflicting keywords in the prompt?

Answer:

There are a few conflicting keywords in the prompt that may pose challenges in incorporating them simultaneously into a single AI-generated image. Here are some examples:

1. Shot on 50mm lens vs. ultra-wide angle: The 50mm lens typically provides a more natural perspective similar to human vision, while an ultra-wide angle lens creates a much wider field of view, often with distortion at the edges. These two effects are difficult to incorporate in the same image.
2. Depth of field vs. hyper-realism and high details: Depth of field creates a selective focus, with blurred areas in the foreground and/or background. In contrast, hyper-realism and high details suggest that the image should have sharpness and clarity throughout. Balancing these effects can be challenging.
3. Moody lighting vs. daytime lighting: Moody lighting typically involves darker, more dramatic, and low-key lighting, whereas daytime lighting suggests bright and natural illumination. Achieving both effects in a single image may not be feasible.
4. Soft lighting vs. harsh: Soft lighting results in diffused, gentle illumination with soft shadows, while harsh lighting creates more intense and direct illumination with sharp shadows. These two effects are opposite and cannot coexist in the same image.
5. Shutter speed 1/1000 vs. F/22: A fast shutter speed of 1/1000 helps capture fast-moving subjects without motion blur, while a small aperture like F/22 increases the depth of field and requires more light. In most cases, a fast shutter speed and a small aperture are not used together due to the amount of light required to properly expose the image.

These conflicting keywords would require trade-offs and prioritization based on the desired final outcome of the AI-generated image.

I found that it’s pretty enough to conclude with the final prompt.

Question: How can we sharpen the first prompt?

To sharpen the first prompt and make it more concise, you can focus on the essential elements and remove redundant or conflicting keywords. Here’s a revised version of the prompt:

“Create an AI-generated image featuring a colonial landscape of Ha Long Bay with ultra-realistic, cinematic photography. The image should be high resolution (8k), with depth of field, beautiful color grading, and moody lighting. Use a 50mm lens perspective and prioritize the composition, detail, and atmosphere. Incorporate ray tracing, global illumination, and shaders for enhanced visual effects.”

This prompt retains the key elements while removing redundancies and conflicting keywords. It emphasizes the desired visual style, subject matter, and technical specifications, providing a clearer direction for the AI-generated image.

___________________

After this conversation, I have a thought about a future in which human-ai interaction will replace human-computer interaction. That may be a distant future (!?), but I find it intriguing and enlightening. As humans, it is more important than ever to remain human, curious, critical, creative, protective of our own nature, and open to change.

In that it can provide answers to questions that we have, I’ve found that AI is a lot like my Tarot cards. Although the Tarot Card’s answer isn’t always accurate, it is ultimately up to us to give it meaning and take responsibility for the choices we make.

But, I mean, isn’t it an incredible time to be alive?