Midjourney produces the most consistently attractive output of the major image generators and has the least useful educational material of any of them, which is an odd combination. The documentation exists, it is accurate, and it is fundamentally a manual rather than a teaching resource. What the docs do well is parameters. Aspect ratio, stylise, chaos, weird, seed, tile, style reference, character reference, the whole set is documented with examples showing what changing a value does.
That comparative approach is the right way to document this, because the effect of stylise at 50 versus 500 is something you have to see rather than read. Anyone confused about what a parameter does will find the answer here. The feature explanations for the more recent additions are decent. Style references, where you point at an image to establish an aesthetic, and character references, where you maintain a consistent subject across generations, are the features that changed what the tool is good for, and the documentation covers them clearly enough to get started.
Office hours are the genuinely unusual part. The team runs regular open sessions where they discuss what they are building, why, and what the constraints are. Hearing the people who built the model explain how it thinks about aesthetics, or why a particular behaviour exists, is more informative than any third party tutorial. It is also unstructured, unindexed and buried in Discord, so extracting value requires effort.
The community showcase teaches more than the documentation does. Seeing what people produce, with prompts visible, is the fastest route to understanding what the tool can do. I have learned more from reading other people's prompts than from any official material. Now the substantial gap.
None of this teaches you to make good images. Prompt guidance amounts to describe your subject, add style words, adjust parameters. That gets you competent generic output, which is what the vast majority of Midjourney work is. The difference between that and something genuinely good is composition, lighting, colour relationships, focal length, framing and a point of view, and the documentation says essentially nothing about any of it.
The people producing excellent work with this tool are almost always people who already had visual training. They prompt for a 35mm lens at f1.8 with rim lighting from camera left because they understand what that produces. Nothing in Midjourney's material tells you that vocabulary exists or why it matters. A course on photographic lighting will improve your Midjourney output more than rereading the parameter docs will.
The Discord dependency is a real barrier for many people. Midjourney has a web interface now and the culture, the office hours and much of the community knowledge remain Discord centred. If you find Discord tedious, and plenty of professionals do, you are cut off from most of the value here. There is no learning path.
No beginner sequence, no progression, no projects. It is a reference organised for lookup, which is fine for a manual and does mean anyone hoping to be taught will need to construct their own curriculum. Practising costs money. There is no meaningful free tier, so working through the documentation requires a subscription.
Ten dollars a month is not much and it is a barrier to casual learning that the free alternatives do not have. My three point four is for accurate, well organised parameter documentation, unusually open communication from the team, and a community that teaches by example, marked down because it is a manual rather than instruction, because it ignores the visual craft that actually determines output quality, and because the Discord centred culture excludes a lot of people who would otherwise engage. Read it in an afternoon, then go learn about lighting.