Midjourney Beginners #5: Things to know and do before jumping into Midjourney V7

A first detailed look at Midjourney V7

A long wait for Midjourney Version 7

Midjourney announced that its V7 would be "coming soon" in late 2024. Then it was a series of "in two weeks, another two weeks, next weeks.." waits before finally launching today, April 4th, 2025 as Midjourney V7 Alpha.

(Read the announcement here)

It’s a highly anticipated launch, given the background buzz surrounding ChatGPT-4o's image creation capability, which has taken social media by storm. Expectations are extremely high for Midjourney to push the boundaries even further. Could V7 meet these lofty expectations?

---

Aesthetics first

David has repeatedly stated during Office Hours that Midjourney prioritized image aesthetics over other things, unlike other AI companies. But beauty is subjective. On one end of the spectrum, some people enjoy images of cute puppies and kittens with an adorable aesthetic style, and on the other, some dudes enjoy horror images with a dark and scary style.

To address this, Midjourney has introduced the Personalization feature in version 6, which allows users to fine-tune the model to generate images that match their personal preferences. In some Office Hours sessions, David even stated that in the future, no one should use the general, default model because the Personalized model improves the Midjourney experience.

So, in V7, Personalization is activated by default. Users are initially prompted to personalize their model. While it is a novel approach to improving user experience in image generation, requiring users to sift through 200 images may be daunting for some. This approach differs from other image generators in that it does not require users to complete any tasks before proceeding, which may surprise beginners who have recently started using Midjourney.

---

Generate many images and choose the best one

Another unconventional approach is to encourage new users to create a large number of draft images and then select the best one. The idea is that it is difficult for users to articulate the type of visual they want, so by telling the model roughly what they want and iteratively improving the images along the way, they can achieve the desired output.

Using Draft Mode, this approach allows for the generation of a large number of options at a low cost and fast. However, users accustomed to GPT-4o's simplicity, which typically generates only one image per request, may notice a change. The additional ideation stage may be a learning curve for beginners, but it provides more control for those willing to experiment. Of course, this is not an issue for intermediate and advanced Midjourney users because we generate a large number of variations and remixes in the older models anyway.

---

The Praise

There’s some debate on Discord about whether Midjourney V7 is better or worse than previous versions. You have to try it for yourself to decide!

Here's my take on the Pros:

• Higher-quality images (even before Upscaling)

• Better aesthetics

• Detailed texture

• Better prompt understanding

Better prompt understanding. The prompt for these images is "extreme macro photography of a leaf". The images in the upper row are the V7 results. The images in the lower row are the V6.1 results. V7 clearly outperforms V6.1 in terms of prompt understanding. (I also tested using different prompts with at least four repetitions per model per test.)

In the later sections, I'll also discuss the "Cons" and how to overcome some of them).

---

Table of Contents

1. Important things to get right before you start

2. Midjourney V7 features discussion

   1. Why Personalization is essential

   2. Image prompting

   3. Draft mode

   4. Style References (Sref)

   5. Moodboard

   6. Compatibilities

---

Important things to get right before you start

These are the three things: (that I wish I knew before starting)