What difficulties arise in working with neural networks?

It would seem that the news about generating images using AI is no longer news - everyone understands how easy and simple it is and rushes to create thousands of illustrations on Midjourney.

But there's a small catch. If you're a marketer in business and hire a contractor to generate images, you'll definitely encounter a pain point - you'll need an art director and an illustrator again.

No, seriously. Right now, our marketing team is preparing new presentations for clients. We want to enrich them with cool images from neural networks. And we immediately encountered the problem of expertise of prompt designers - they are good at managing the neural network, but they cannot envision the final result themselves.

So, a dry marketing brief won't work for them. They need a highly detailed creative brief describing all the details: background, clothing, elements, etc. As you understand, this is usually done by an art director.

And it would seem - we replaced the illustrator who draws with the prompt designer who programs the neural network. And again, it's not all rosy - if you're making a series of images, you start dancing with tambourines and with the neural network to make it produce identical faces in all the pictures, adhere to a unified design code for interiors, or even just the number of pens on the desk. In our case, the illustrator has to finalize the images to the final result.

Nevertheless, AI certainly speeds up the process and makes creative work faster. Our post-production team is actively exploring prompts in the same Midjourney.