Sure things. I've been using Midjourney v5 for a few weeks, and there pictures below was made by it. Also, I only provided prompts without any post-production.
I would like to share some of my points with you all if you are interested in.
After experiencing MJ v5 for a while, let me summarize the biggest impressions.
There are two points.
Firstly, I believe those who have used it a few times would have the same feeling: at the current stage, it is difficult for users to control the scene very precisely through prompt words. It can be said that the result is in a vague state within the given prompts range. Just give AI more time.
Based on the understanding of the first point, the second point arises. How to obtain a scene that meets one's needs as much as possible? At present, it still lies in the prompts themselves.
From the view of the composition of the scene elements, the prompts should pay special attention to the overall logicality and accurate specificity. The latter is easy to understand, as the prompt words provide specific visible objects or scenarios. For example, you may say you want a romantic scene, but what is a 'romantic' scene? How does AI understand 'romantic'? Does it mean the same as what most people understand by 'romantic'? Instead of giving the prompt 'romantic' directly, I would recommend describing something like this: a French girl dancing in a forest under the backlight during the golden hour. This indeed requires imagination.
Then what does the logicality of the scene mean? I found that the more detailed the prompt words, the better they are not. I think AI's processing method is trying to give you what you need. However, the prompt words you give may be contradictory to AI. For example, you give "close-up shots, out-of-focus light spots" and a specific scene content "empty square". In my opinion, humans can understand these three descriptions, but putting them together for AI is contradictory because, logically, to shoot a close-up shot with out-of-focus light spots, environmental information is extremely lacking. However, you also provided AI with environmental information. In this process, it has to choose between the two, and the final result might be a close-up shot without knowing where it is, or a scene in the square, but not a close-up shot.
Please feel free to talk about this topic!