Can Midjourney AI Perfectly Recreate Your Pet?
An investigation of version 5’s image generation techniques
--
Now that Midjourney AI version 5 has been released, I decided to put it to the test. I aimed to try a few different methods to recreate Grizz, my mixed-breed rescue dog and longtime best friend. I tried this experiment with Midjourney version 4 and wanted to see how version 5 improved.
Test #1 — Single Image + Text Prompt
For the first test, I uploaded a picture of Grizz (above left) and used the image plus a text prompt to generate an AI version of Grizz (above right).
What I instructed Midjourney to imagine included both the image of Grizz and a text prompt to try and guide Midjourney AI to my desired result.
Text prompt:
The good dog named Grizz is looking up toward the camera
Although the coloring, collar, and some physical features of Grizz were slightly off, I was impressed at how real the generated image looked. I did not include much detail in the written prompt, but what I did was considered. The camera angle and focus reflect the words of the prompt. I could be crazy, but I feel like the essence of Grizz was captured. The look on the generated image’s face feels real.
Test #2 — Triple Image
For the next test, I wanted to see if adding more images to the prompt would help Midjourney create Grizz. I snapped a few pictures of Grizz and was ready to try it. I chose to take similar pictures to give Midjourney the best chance to understand what I was looking for.
I performed the test with version 4 and version 5.
Text prompt:
Grizz
Let’s analyze the results. Both versions 4 (above left) and 5 (above right) produced good images, but the versions prioritize different…
