AI Art Adventures

So after having compared Dall-E and Midjourney, Dall-E is easier to use by far and can generally kind of follow the prompts you give it. The art style is definitely simpler/less artistic. I guess it pulls from a different database to build its images.

You also definitely need some sort of photo editor to finish out the images (I don’t have one and couldn’t figure out Photopea,) but the bots are definitely getting smart enough to get you within 80-95% of what you were actually imagining.

Dall-E: “white silhouette on black background, skull with lightning bolt in teeth, smooth, 4K, by artgerm” (best two out of four, none of which were horrible)

Neither of these include a lightning bolt where it was actually wanted, but these are pretty darned close and of just about the appropriate quality. I wanted something that I could plausibly print on a patch for my jacket. (cough.) The upgrade function on this program isn’t great, though, as subsequent variations got worse and worse instead of better.

Dall-E also has a partial edit function which allows you–if you phrase things carefully enough and the bot already has a clear idea of what the image contains–to expand and alter your image. So I could complete that one skull.

And with quite a bit of work a few weeks ago, I also managed to produce:

Midjourney, on the other hand, is a royal PITA to use if a) you don’t know how to use Discord, b) you don’t pay for the subscription allowing you to direct-message the bot. It does much better with the more artistic type digital art renders, but when I fed the same exact skull/lightning prompt to it this morning, I got these:

Now, yes, 2 and 3 aren’t bad and can be worked on, but neither of those are going on a morale patch any time soon.

On the other hand, though, “pencil sketch, knight in plate armor and long cloak holding a sword overhead, ultra detailed, realistic”

The best of Dall-E’s four variants is on the right. The best of Midjourney’s is on the left, and isn’t it just kind of awesome??

Anyhow, speaking of 85%:


7 thoughts on “AI Art Adventures

    1. There’s definitely a (hah) art to getting what you want out of these things. Some people have managed some really awesome things and the algorithms are learning.

      I for one welcome our new all-digital moviemaker overlords…

      Liked by 2 people

    1. Yes, yes, I have no way of knowing but Midjourney wouldn’t let me use the word “bleeding” in a prompt. So they at least are keeping up legal plausible deniability.

      I think that the technology benefits the semi-talented, the people who have an image in their head but lack the ability to get it out of there whole.

      Liked by 1 person

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s