Mastering Midjourney: A Food Illustrator's Personal Journey into Neural Networks

Mastering Midjourney: A Food Illustrator's Personal Journey into Neural Networks

Anna Makarova
May 15, 2023
Mentioned in this story

picture generated in Midjourney by the author

It's common these days to debunk paid courses. Maybe they're not all that bad. But what's definitely worth debunking are the neural networking courses. Where would professors come from all of a sudden when the topic is a year old at most? Anyone who dives into studying it persistently can catch up with the smartest people in a couple of weeks.

I've been working as a food illustrator for ten years. And now I'm mastering Midjourney and trying to develop my expert intuition here as well. That's what seems important to me personally:

  1. Don't seek immediate benefits
  2. Learn with a community
  3. Change your way of thinking
  4. Focus

1. Don't seek immediate benefits

Probably, many of you have encountered people who are fixated on the monetization, commercial potential, and practical application of neural networks. They are intensely pondering how to immediately put AI to work for them and finally start recouping investments and generating profit.

Sometimes, some of my neural creations get sold, but I still believe that it's not worth making a serious bet on it right away. Otherwise, you might sacrifice that special mood that arouses curiosity and ultimately develops skills.

It's too early to worry that no one wants to buy your sand pies, it's better to learn how to make them magnificent for now!

Have fun, play, create art.

Corporations and brands are still watching closely, a paradigm shift is needed, breaking of templates, recognition of a new aesthetic. Few can resist counting fingers* in neural images, so the general suspicion needs to be calmed with beauty.

2.Learn with a community

The simplest learning event is a pair of 'picture+prompt' together. Like a key with a lock, or a problem with an answer. The brain perceives both elements in conjunction and learns something (you don't always understand exactly what). There are no other learning events here. You need to feed your brain with these things. You create some of them for yourself, but you still need more!

Since there are no universities for prompting, and most likely, there will not be soon, learning through a community comes to the forefront.

We were dropped more or less simultaneously by parachute into the same forest, and everyone is now digging their own clearing. But if we don't get to our people, there's not much chance of success. The challenge has been thrown down by a superior mind. We can only resist it if we become a swarm.

Communicating in thematic chats and similar places (they have now appeared in all social networks) is not only entertainment but also a necessity. The exchange of information there happens almost at the speed of thought. The power of the community lies in the fact that it is an environment for the evolution of ideas. You will get back the time spent in chat, saving it later on your generations.

In enthusiast gatherings, trends are born and die, words of the week or month emerge. You can observe this from the side if you don't want to get involved. Some words indeed turn out to be more useful than others, and since a multitude of examples for each one appear immediately, generalization occurs. For example, 'humanization', 'balenciaga', 'golden hour', 'knitted', 'intricate', 'greg rutkowski', 'ivan bilibin'...

In a community, you can gather pollen; alone, you will continue to chew silage.

3. Change your way of thinking

The previous points on the list might have seemed obvious, I hope this main one will not seem difficult now.

A good contact with a neural network requires operating with some new concepts in your mind. For example, an experienced prompt engineer thinks in sets. A specific result of any generation - four pictures - is also a portal, which leads to the whole mass of generations caused by the same prompt. And you have to sense them immediately. We weren't taught that, it's a meta-level, but now it's a must.

It will probably be useful to acknowledge that your mental abilities are quite limited and think about how and where to compensate for this.

A specific example: for the neural network, English is the native language, but for many users, it's not. It's impossible to quickly learn English to the maximum level, but you can cover weak spots. Pay attention to word order, articles, run the text in different versions. Sometimes it's boring, but you won't miss the important. Here are two similar requests: 'food for dogs' and 'dog's food', but the difference is fundamental.

A humble illustration: I often use ChatGPT to get a list of synonyms.

Such tools need to be sought and developed, we can't do without them, a weak individual mind needs an exoskeleton.

4. Focus

You need to focus in order to save time and attention, otherwise you'll lose out on both. If you plan to move from a superficial and amateur level of working with neural networks to an expert level, you have to narrow down the field one way or another.

Something like religions has already emerged: some believe in long Stable Diffusion prompts, some in short Midjourney ones. I'm in the sect of short ones. Besides, I once limited myself to the theme of food in illustration, and now it's neuro-food.

The main choice is subordinated to smaller choices: what to read, what to watch, which of my colleagues to communicate with. For example, instead of channels-aggregators, which still use the ridiculous formulation "neural network created", I prefer to read the authors', in which there is a process of search, published promos, there is a live chat.

Tip: If you work on your own server in Discord (recommended), make separate channels there, spread different projects over them. For example, this is how I collect other people's Prompts, and they pile up as a separate bank of ideas.

All written above was about saving attention.

It is very important to save time as well.

Install an app on your phone, and send the generation spinning any time an interesting thought visits. This approach yields a great deal. To become a prompt pro, you need to generate, generate, generate. Do it in the elevator, in the store, in the check-in line, while standing in traffic, or waiting for a hand at poker.

Colleagues have switched to shortcuts in MJ, this is already a reality. If you generate a lot, you also have to optimize everything. Generating little will not work!

Mentioned in this story

link copied