If you’ve been online for a while now, it’s very likely that you already know about Artificial Intelligence and how it’s changing the game in pretty much every field. I recently started to explore the new Version 5 of Midjourney (a tool that generates images from textual prompts) and I fed it some of our editorials throughout these past few years and we were blown away by the overall quality of it.
This new version of the product provides a higher level of detail, including some areas where previous versions were not very accurate, such as faces and hands as you can see in the pictures above. It wasn’t unusual for some AI pictures to have people with weird-looking eyes and hands with 12 fingers.
With this new version, it’s much easier for anyone using Midjourney to obtain a more personal aesthetic using a new approach to prompts. A cited by the author of the article: “When writing prompts… forget the concept of separating words with commas and slashes, with parameters and style methods. Rather, write it all in a sentence“.
The richness of the final result depends mostly on how you word your prompts, what are some specific details that you want to incorporate into the image, what’s your preferred aesthetic, and so on. In our case, we used some of our own images for reference by adding them into the prompt. This is done by uploading an image to Discord, and then copying the discord URL of that image into the prompt after /imagine. That would result in something that looks like this:
That prompt generated this image:
There are two important aspects of this new version of Midjourney that called my attention: the fact that images are generated in much more higher resolution from the beginning and are upscaled faster to 1024×1024 resolution. I’ve noticed that depending on the complexity of the prompts and the amount of reference images that I use it can take longer to process, but the results very rarely disappoint.
Be aware that certain prompts are banned, and certain types of images are not allowed as well. If you’d like to see the list of banned terms and types of images, as well as other ToS, Midjourney has a resource here that you can check.
I’ve had a few circumstances in which I got a message that my images or prompts were not complying with the Community Guidelines and it’s been mostly because an image was identified as something potentially NSFW, or one of the prompts could’ve be interpreted as circumventing the Terms of Service. This is normal and it’s likely that it can happen to you at the beginning as you start to explore Midjourney, however, if you insist you can get your access permanently revoked and you’d be unable to use the platform again.
Right now Midjourney has a range of subscription plans, including a free trial so you can test the waters and play around with it. The prices go from $10 all the way up $60 including commercial terms.
Will AI fully take over modeling jobs? In my opinion it is highly unlikely, even though there are a few applications in which they’re already being used, such as e-commerce, digital advertising or catalogues.
One of these AI-model generators is Lalaland, that aims to challenge traditional methods for creating e-commerce catalogues, by facilitating the process of adding products in bulk, allowing their clients to customize photos and models physical appearance to align more closely to their business and communication goals to ensure that there’s more diversity in their products in a more cost-effective and sustainable manner.
I believe these tools will definitely impact the way we are accustomed to consume digital products and there will new ethical, political and financial discussions that will need to take place, but it’s still very soon to foresee how AI will impact our lives directly in the near and long term future. I do believe that right now is the best time to learn how to use AI tools to enrich our creative exploration and complement tasks in our daily lives in different fields.
Something is certain right now, AI models will not be walking runways anytime soon.