✷ why I work with dall-e instead of midjourney
As an artist, I enjoy using AI tools to produce more work; however, I almost exclusively use DALL-E rather than the other image generation tools. People have asked me why I don’t use Midjourney, so here are my feelings about these tools, as someone who uses them.
Working with DALL-E is like being an art teacher working with a brilliant pupil. This pupil has done a lot of research and a lot of work on their own, and knows exactly what “pre-raphaelite art” looks like or what paintings by “sir john everett millais” tended to have in common, but simply struggles with a few basic things like making hands, eyes, and faces look accurate rather than monstrous. In the role of an art teacher, it’s easy to cluck your tongue and say, “oh, DALL-E. Here, let me step in and help you,” when it comes to fixing mistakes. DALL-E’s work almost always needs significant editing in order to be a satisfactory rendition of what I had in mind---it rarely produces something satisfactory on the first try. But this is a good thing. To me, as a human artist, this feels encouraging, because there is a clear role for me in the process. I am the one guiding DALL-E, I am the one looking over its shoulder and fixing its mistakes. Sometimes I do this by hand, and sometimes I use the technique called inpainting, where I erase a certain portion of an image and ask DALL-E to fix its own mistake.
Sometimes it can’t---sometimes DALL-E just doesn’t know how to produce what I want (e.g. hands that look good, grass that is a natural color or texture and doesn’t have streaks of fluorescent colors in it, eyes that resemble normal eyes). But as a human, I like that, because it means I can be more involved in the process, since I am involved in fixing DALL-E’s mistakes. The final product is my work, and I’m proud of it, because DALL-E couldn’t have generated something this sophisticated and reasonable-looking without my help.
Midjourney, on the other hand, seems to leave little room for me as a human artist. It is perfectly capable of producing good, satisfactory work the first time I give it a prompt, which intimidates me. It does not need me to go over its mistakes the way DALL-E does. The process is so much less involved. My experience with Midjourney is from a generation that I made where I asked it to produce a “pre-Raphaelite squirrel.” On the first try, two of its four prompts would have easily satisfied my standards for realistic, stylistically appealing art with no editing whatsoever---whereas with DALL-E, that standard is often met by one of the generations only after the eighth attempt. So I react to Midjourney the way I would to a new competitor, who has surpassed my own artistic abilities – with instinctive feelings of insecurity and intimidation. Some new rival has entered my artistic domain! And they don’t even need to spend time erasing and reworking to account for their mistakes, their work is finished on the first go!
Maybe I've been too harsh on Midjourney. There are artistic ideas I have come up with that DALL-E seems simply unable to execute well, like “kitten archaeologists," which Midjourney may have better luck with. Having grown frustrated with DALL-E’s attempts, which were nothing like what I wanted stylistically or in terms of accuracy, I made my own pencil drawing to satisfy the urge to get this image onto paper. I like DALL-E because it inspires me to do more, to make more, to be more involved in art. DALL-E simply saves time, it does not replace my own skills. It simply isn’t capable of depicting things that I’m perfectly capable of drawing, like kitten archaeologists. But with Midjourney, Midjourney’s compositional skills do surpass mine, especially for larger pieces. Whereas I feel DALL-E is simply a time-saving shortcut, or a collaborative but often-misguided partner in the creative process, Midjourney is an entity whose artistic skills are farther beyond than what I imagine myself being able to reasonably accomplish, and that is what makes me feel intimidated by it.