So, last time I started with outpainting, close ups, fashion, gelled lighting, black and white portraits. I specified 10mm, but it didn’t really get fisheye like. Maybe I need to do both, and specify fish eye? Also what happens if I add ‘reaching towards camera’?

Studio close up of a beautiful 20 year old woman reaching towards camera, black and white, Sigma ART 10mm f/1.4 100iso, dramatic lighting, fish eye lens

I really, really hoped this image would include a hand reaching to the camera, but its definitely more fish eye. I don’t actually think the focal length gives it a look, like at all. It might be helping find professionally tiered images.

Photo of slim 20 year old girl, LEDs visor helmet, profile pose, high detail, studio, smoke, cyberpunk, 85mm Sigma Art lens, in the style of Martin Schoeller

I am disappointed to see almost no Martin Schoeller influence here.

Studio portrait of beautiful 20 year old woman, machine cables coming out of neck, eyes shut, dramatic red light, Sigma 105mm F1.4 100 iso, fog

I was hoping to get something a little more matrixy out of this, but these are interesting.

Rewording my prompt to: Studio portrait of beautiful 20 year old cybernetic woman , eyes shut, dramatic red light, Sigma 105mm F1.4 100 iso, fog, david bowie inspired makeup

I really hope the David bowie makeup pulls through.

I am really loving this, but I don’t see any David bowie in it, except the models became more androgenic. And no makeup that I see.

Studio portrait of beautiful 20 year old woman touching her face, profile shot, high key butterfly lighting, Sigma 105mm F1.4 100 iso, david bowie inspired makeup

Loving it. It doesn’t know what butterfly lighting is, so I’ll have to work on this.

Studio portrait of beautiful 20 year old woman looking straight into camera, single light from above, shadow under nose, Sigma 105mm F1.4 100 iso, david bowie inspired makeup

It’s not quite getting the lighting how I want it (just to remind you, this is butterfly lighting, so if you know how to describe that, send me an email please!) Makeup is getting more interesting.

31 credits left as of writing, still a lot to go. Let’s see how it handles locations.

Beautiful 20 year old woman smoking a cigarette in a bathroom, on-camera flash, Sigma 105mm F/2.8

Hmm, not quite. I think I want them a bit more dishevelled, and more dramatic lighting to invoke Nan Goldin.

dishevelled 20 year old woman laying on a bed smoking a cigarette in a bedroom, dramatic lighting, Sigma 105mm F/2.

Interesting. Almost Nan goldin, but it feels snapshot to me.

Photograph of a beautiful 20 year old woman, fashion photography, wet hair, rain, standing on a rooftop, cyberpunk city, futuristic city, sigma 105mm, f/1.2, 100 iso

Starting to not be good in the face, though I love the third one, minus the woman.

but let’s try a few more locations just to see how it handles landscapes without people.

Landscape photo of a foggy field with a single tree two thirds from the left.

I want to make a castle similar to zelda.

Landscape photograph, Sigma 30mm F/11 100 iso, gothic castle, iron age, mystical, mythical, magical, clear skies, sunlight, clifftop, holy light, hyper realistic, kingdom, fairy, earthly, 4k, highly detailed, peaceful, beautiful, Hyrule

Okay, so Midjourney is much more fantasy/epic than Dalle, good to know.

Studio portrait of beautiful pale 20 year old woman, avant-garde fashion, high key lighting, Sigma 105mm F1.4 100 iso, white outfit, white background

Oh boy, it did NOT like avant garde.

Studio portrait of beautiful pale 20 year old woman, vogue fashion, high key lighting, Sigma 105mm F1.4 100 iso, white outfit, white background

Okay, kinda interesting but that’s more over exposed than high key. Also, I want a more eccentric outfit, but I’ll forget about that for now. Also sticking wide angle just to see what happens.

Editorial vogue fashion of beautiful 20-year-old woman, fashion, dramatic low key lighting, Sigma 105mm F1.4 100 iso, dark outfit, dark background, wide angle lens

(though I realized I used fashion twice)

I am loving the fashion, especially in the last two, and vogue definitely gives it more interesting poses. The second one is far too bright, and the first one has a bad face.

I want to see how the AI handles backlighting, so I am going to modify an old prompt I liked.

Photo of a slim girl, 20 years old, side profile, close up, high detail, studio, Back lighting, grunge fashion, smoking a cigarette, cigarette smoke, Sigma 105mm F/2.8

I love shots of Lindsay Adler, especially when she does close up of lips with beauty lighting. Let’s see what that gets me.

Extreme close up of female lips, 20 year old woman, bright red lipstick, beauty lighting, studio, slight snarl, teeth, Sigma 105mm, F/1.8, 100 iso

I’m not really sure what I expected, but this was not a successful vision. I wanted a snarl kinda like David Bowie does.

Let’s see if I can shape the light, Adler style.

Studio portrait of beautiful 20 year old woman, eyes shut, narrow cheekbones, narrow light in the shape of a square on the subject’s eyes, Sigma 105mm F1.4 100 iso, fog

Not what I pictured, but cool. I would love to have the face under exposed, but exposed properly on the face.

Studio portrait of beautiful 20 year old woman, narrow cheekbones, ponytail,  face is underexposed, eyes are lit, Sigma 105mm F1.4 100 iso, fog

It’s really struggling to understand me. These look poorly lit because I specified under exposed, but I wanted a blotch of light hitting the face, cut with cookies.

Studio portrait of beautiful 20 year old woman, narrow cheekbones, light pink hair, pink background, pink makeup, pink dress, narrow light in the shape of a square on the subject’s eyes, Sigma 105mm F1.4 100 iso, fog, analogous colour scheme 

(I discovered it can do schemes in the makeup section, I’ve been writing this out of order).

It’s funny, I don’t think I changed too much from the prompt from earlier, but it really removed the dramatic/low key lighting.

I want to do something like medusa before I run out of credits, so let me try this…

Studio portrait of beautiful 20 year old woman, narrow cheekbones, snakes for hair, pink background, pink makeup, pink dress, narrow light in the shape of a square on the subject’s eyes, Sigma 105mm F1.4 100 iso, fog, analogous colour scheme, dramatic lighting

I’m surprised it ignores snakes entirely.

Editorial fashion photography, Beautiful 20 year old woman in a reflective outfit, LEDs visor helmet, blue glowing orb hovering over her outreached hand, white studio, Sigma 105mm F1.4 100 iso

The orb didn’t work too good.

Editorial fashion photography, Beautiful 20 year old woman in a reflective outfit, LEDs visor helmet, dramatic red light, studio, Sigma 105mm F1.4 100 iso, vogue fashion

Editorial fashion photography, Beautiful 20 year old woman in a reflective outfit, hand painted background, pastel light, studio, Sigma 105mm F1.4 100 iso, vogue fashion, hands outreached

Well, that lost the drama.

Editorial fashion photography, Beautiful 20 year old woman in a reflective outfit, dramatic light, studio, Sigma 105mm F1.4 100 iso, vogue fashion, hands outreached, fish eye lens, muted tones

Interesting how the fish eye mostly (mostly) accurately warps the body.

Editorial fashion photography, Beautiful 20 year old woman, dramatic light, studio, Sigma 105mm F1.4 100 iso, grunge fashion, hands outreached, muted tones

Final one: Beautiful 20 year old woman, short hairstyle, sitting beside window drinking wine, Sigma 105mm F1.4 100 iso

I don’t know why, but this prompt was really repetitive. I guess I should have expected that.

See you in part 4: cosmetics and food.

I’ve been really, really looking forward to Dalle-2. As of writing this, there is a waitlist (no invitations, they manually do approve people). I applied July 6, and it took over a month, being approved on August 10th. The AI was trained on millions of images. It knows photographers, painters, camera settings. The key to it is to be specific in your prompts, and from that, the AI takes over. Since AI is computationally intensive, and you only have 50 free credits, each prompt needed to count. The limit does make it somewhat daunting. Keywords like “medium shot” “close up” “long shot” seem to determine how far the camera is, and you can also specify camera settings. You can also use this clip software to *kind of* see if the AI might know what you’re talking about. For instance, I can see it does have entries for Nick Knight and paolo roversi but not Lindsey Adler, but that isn’t perfect. You also don’t want to prompt with negatives, as that can confuse the AI.

A high quality full body editorial fashion image of a redheaded woman looking slightly left wearing a red dress, key light to the left, Sigma 105mm F1.4 100 iso, white background, in the style of Nick Knight

Although it is a beautiful dress, the images that do have a face look very not realistic. I however, love the dresses and think they’re really good. The one on the right looks like it could be really in vogue. 

Head and shoulder portrait of a red-headed woman wearing a red shirt in a studio, Sigma 105mm F1.4 100 ISO, Rembrandt lighting

The subjects aren’t particularly attractive, but it sure looks real. So, how can I get a full body image that includes the face? I’m wondering if the AI only saw ‘full body’ and thought it didn’t want a face. But otherwise, it’s starting to come together, and I love the lighting, particularly on the left.

A long shot studio photographic portrait of a beautiful red headed woman, Sigma 105mm F2.8 100 ISO, in the style of Paolo Roversi.

Okay, I’m kinda liking this, but I don’t see much Paolo in it. Paolo also does a lot of nudes, so maybe I should steer clear of this, as the creators were very careful about what they submitted to feed it. Also, wondering why it’s not a long shot. But I like where this is going.

A full shot studio editorial fashion of a beautiful woman wearing grunge clothing, Sigma 105mm, dramatic lighting

I am really, really into the outfits, but I am not sure where the faces went. It seems to do bodies OR faces.

Head and shoulder portrait of a beautiful woman wearing sunglasses with neon signs reflected in them, in an alley of a futuristic city, night time, Sigma 105mm f/2.8 100 ISO

I wanted to see what would happen if I changed from f2.8 to f/2.8.

These look a bit less professional but very real, but still kind of cool. It feels ‘snapshot like’ like someone took this image of their friends.

Head and shoulder studio portrait with red and blue studio split lighting of a beautiful woman in grunge clothing, blowing bubblegum, sigma 105mm F/2.8 100 iso, purple background.

I was going for Jill Greenberg here. Gotta say, I’m making some progress. I just can’t figure out how to get the full body in the shot, but I think I will need to do what is called outpainting (I’ll come back to this).

I am going to a keyword-based approach.

photo of slim girl, 20 years old, close-up, high detail, studio, smoke, sharp, pink violet light, studio, 105mm sigma art lens

I like how the fog looks, and the lighting is starting to add drama. There are nice shadows on the faces.

Photo of a slim girl, 20 years old, side profile, close up, high detail, studio, dramatic lighting, grunge fashion, smoking a cigarette, Sigma 105mm F/2.8

Okay, so this is starting to get a little more interesting. We can actually shrink the image in photoshop, put transparency around it, then Dall-e should be able to fill in the rest, this is what openAI calls what is essentially uncropping. Let’s see if it works.

I ran the same prompt again: Photo of a slim girl, 20 years old, side profile, close up, high detail, studio, dramatic lighting, grunge fashion, smoking a cigarette, Sigma 105mm F/2.8

and it gave me:

So, this got much less realistic. I’m going to try repainting one of my earlier grunge images that I liked the body of, and see if it does a better job getting a head from a body. Maybe I should have removed close up?

Photo of a slim girl, 20 years old, high detail, studio, dramatic lighting, grunge fashion, Sigma 105mm, sharp

After talking to people I learned that outpainting is better if you erase less, and don’t make the original too small, so I am doing one last try: A full shot studio editorial fashion of a beautiful woman wearing grunge clothing, Sigma 105mm, dramatic lighting

I’m done with outpainting, haha. Doesn’t matter how little I zoom out. It is awful. Need to look more into this.

Studio close up of a beautiful 20 year old woman, black and white, Sigma ART 10mm f/1.4 100iso, dramatic lighting

Not really fish eye, but definitely dramatic, and definitely interesting results. Onto part 3!

 I have been wanting to journal my entry into AI artwork, and now, I finally can. In school, we journaled a lot. It was helpful to reflect on what we made and keep notes if we wanted to refer back to our thoughts.

But where do I start? Dalle-2 vs Midjourney?

I wanted to start with Midjourney, as there are only 25 minutes of rendering time.

My first prompt was: /imagine prompt: photorealism of full body editorial fashion In the style of Nick Knight :: red headed model fisheye lens –no defocus –hd –quality 2

I am wondering if I made a mistake and mistook photo realism for something it is not.

I changed my prompt to: /imagine prompt: photograph of Grunge fashion photorealistic dramatic lighting –no defocus –hd –quality 2 –ar 2:3

I guess that is kind of grungey? But it’s very cartoony. I am quickly burning through my free minutes, but I wanted to remove photorealistic and switch to landscapes.

/imagine prompt: photograph of vibrant Landscape of an alien planet, purple grass and dark thunderstorms in the horizon –no defocus –hd –quality 2 –ar 2:3

I wish I remembered to change the resolution from 2:3 to 3:2, but otherwise, it looks pretty baren. It’s also not very vibrant.

/imagine prompt: photograph of A woman looking out a high window to see a very detailed science fiction city with lots of neon lights –no defocus –hd –quality 2 –ar 16:9

This looks a lot less realistic than what I’m seeing, so I decided to copy a prompt instead of trying to make my own (modifying it to my own vision)

/imagine city of the future, many skyscrapers, blade runner style, hyper-realistic, octane render, realistic, real, street, rain, beautiful girl standing on a rooftop, pink hair, cinematic, 8k, very intricate, 80’s, night

Interesting. It seems to respond to less sentence, more keywords.

/imagine white studio, hyper-realistic, octane render, realistic, real, street, rain, close up of woman’s face, brown hair, cinematic, 8k, very intricate, 80’s, night

Closer. I was hoping for a wet in studio look.

/imagine prompt:white studio, hyper-realistic, octane render, realistic, real, grunge woman’s fashion, brown hair, cinematic, 8k, very intricate

I am running low on minutes, so I’m just gonna try something.

/imagine prompt:ethereal, floating woman holding a book, spirit, detailed cloak, magic, 8k, very detailed, 

I really like the bottom right. Midjourney has an upscale feature, which I haven’t tried yet. I can make 4 images left, let’s see if that will upscale.

forgotten ruins, highly detailed, sorrow, ugly –ar 2:3 –quality 2

Check back in for Dalle-2 in part 2.