The Power of AI in Image Generation with Midjourney

The Power of AI in Image Generation with Midjourney

In this AI4Marketing episode we discuss our experience testing image-generating AI tools, including Dall-E-2, Midjourney, and Stable diffusion. We continued our test with Midjourney for our podcast episodes. We found out that generating very specific images could be a bit challenging.

We have experimented with different prompts and learned that modifier words and magic words can make a difference in the quality of the generated images.

We think that Midjourney may need more data to generate certain types of images accurately. The prompts that we have used have resulted in our podcast image for episode 1, 2 and 3.

All prompts from this episode:

Prompt 1:
A friendly looking robot that uses typewriter. The robot has already written hundreds of papers.
Comment:
Very nice looking image. But no papers and the robot is one with the typewriter.

Prompt 2:
A friendly looking robot in front of a typewriter. stacks of paper in the background.
Comment:
Some paper, but not a lot. In some images the robot and typewriter are still one.

Prompt 3:
A friendly looking robot. A typewriter. Multiple stacks of A4 size paper in the background.
Comment:
The paper part is OK. But the robot and typewriter are still one thing.

Prompt 4:
A friendly looking robot. A typewriter that is not attached to the robot. Multiple stacks of A4 size paper in the background.
Comment:
Problem still not fixed an the stacks of paper are gone now.

Prompt 5:
A friendly looking robot writing a novel with a pen and paper. Multiple stacks of A4 size paper in the background.
Comments:
I like this one.

Prompt 6:
A friendly looking robot writing a novel with a pen and paper. Multiple stacks of A4 size paper in the background. Studio lighting, shallow depth of field, professional photograph. Pixar style.
Comments:
Don’t like the depth.

Prompt 7:
A friendly looking robot writing a novel with a pen and paper. Multiple stacks of A4 size paper in the background. Studio lighting, high resolution, professional photograph. Pixar style.
Comments:
This one is perfect!

U1 (upscale) to get the right size.

Episode 1:
A friendly looking robot, walking on the street, reading a map. Studio lighting, high resolution, professional photograph. Polymer Clay style.

Episode 2:
A friendly looking robot. Question marks floating in the background, Studio lighting, high resolution, professional photograph. Pixar style.