I wanted to try to combine GPT-3 and Midjourney to impersonate as a pro prompt photographer and create realistic photos that could be mistaken for being real photos taken by a human.
So I got advice on camera, camera settings, lighting and aspect ratios from GPT-3 and used them to create advanced prompts in Midjourney
YouTube:
How did I get advice from GPT-3?
After instructing the AI that it was a professional photographer I simply used the following commands in GPT-3:
How do I take a good photo of a cityscape?
The advice i got for a Cityscape Photo was:
DSLR camera, Canon EF-S 10-18mm f/4.5-5.6 IS STM Lens, ISO 150, shutter speed 1/125s
How do I take a good profile photo of a person?
The advice i got for a profile photoshoot, was:
DSLR, 50mm lens f/2.8, shutter speed 1/60, soft light source
What were the objectives?
Then i took these advice from the GTP-3 AI over to Midjourney and created a prompt to try to pretend that i was going to take a real photo of my objectives that was:
- Outdoor winter shot of The White Walkers from Game of Thrones
- A portrait shot of a Daenerys Targaryen
- A cityscape of the Lost city of Atlantis
- A Wildlife shot of Vecna from Stranger Things
The Results
White Walkers from Game of Thrones:
This was scary good. I really like how realistic the image turned out, the lighting is perfect and the White Walkers look spot on.
So really impressed how this did turn out.
A portrait shot of a Daenerys Targaryen:
WOW! This just blew my mind. Midjourney has really improved faces in this new version. It is almost too good, scary, similar to both Emilia Clarke and Daenerys Targaryen.
I was so impressed by this one, my favorite from this session.
A cityscape of the Lost city of Atlantis:
This one also turned out great. I like the underwater look and feel to it. The city looks like it could be from the lost city of Atlantis. The details are just stunning.
Vecna from Stranger Things:
This one also turned out great. I like how Vecna looks in this image, and the Stranger Things feel to it. The lighting and colors are perfect. But the camera settings did not really work here.
Conclusion:
My testing so far, as you can see in the video, is just positive. The results are just amazing. This is a new level of AI Art generation, and it blows Dall-E 2 out of the water if you ask me.
Can Midjourney impersonate as a pro photographer?
Yes, I think it can in some scenarios, but we will have to wait and see.
So I see this as a big step forward for Generative AI art yet again. And let’s see where this is going when the new Midjourney version 4 is live.