Creating AI Images from Text

There have been several blog posts recently about the use of Artifical Intelligence (AI) for writing school papers, blog posts, or other articles. Hugh Roberts published a great post yesterday about the AI blocks recently added to WordPress. I recommend all WordPress users read it.

If you are a tech lover, I cannot imagine you do not find this interesting. We all love new technology, don’t we? The artist in me is intrigued. Just what could an AI bot produce?

My Disclaimer: My experiments required a bit of bumbling. The AI bot I found ran on Discord. I have of course heard about Discord but had never used it. In the most basic explanation, it is a social forum-like chat platform used by gamers, developers, and people who share common interests. It has a user base that exceeds over 250 million users. In order to use the AI bot you must have a Discord account. As you can imagine, I was bumbling around Discord and also bumbling around the bot.

The bot I used is called MidJourney. It is a text to image AI, meaning you supply it with text and it generates images based on your text. The images produced are referred to as assets. Unless you are a paid member, you do not own the assets you create but are granted rights to use the images. This is a very slippery slope as these images are derived images from other sources. Copyright could come into play. Here is the statement from the MidJourney user guide:

“If You are not a Paid Member, You don’t own the Assets You create. Instead, Midjourney grants You a license to the Assets under the Creative Commons Noncommercial 4.0 Attribution International License (the “Asset License”).

The full text is accessible as of the Effective Date here:”

I would recommend reading through the documentation at great length before you publish AI generated images. I will site this license below on the images I created.

So how does it work? First you login to Discord and then to the MidJourney server.

From there, you supply text for the image you hope to create using an imagine command.

/imagine a small boy with red hair standing alone in a green field with dark skies and lightening storm approaching

The AI bot will generate four images and you can further refine these images. This is the one I ended up keeping.

Used under Creative Commons Noncommercial 4.0 Attribution International License

My second attempt

/imagine an ethereal scene with a realistic young girl with shoulder length windswept brown hair dressed in a lilac colored long gown riding a unicorn with pale pink wings against a pale blue sky

Notice that there are two horns on the first unicorn and only one wing. One the second image, the wing seems attached to the girl’s head rather than the unicorn and the proportions are definitely off. Was my description misleading so the bot thought I wanted the girl to have wings instead of the unicorn?

Used Under Creative Commons Noncommercial 4.0 Attribution International License
Used under Creative Commons Noncommercial 4.0 Attribution International License

The last one I will show I did manage to keep both the first four images the bot created and the final image (after a few generations) I kept.

/imagine a tall slender woman with long windblown blonde hair wearing denim blue jeans and a plaid shirt and cowboy boots standing beside a red truck on a dirt road in the deserted countryside, blue skies with white clouds

Used under Creative Commons Noncommercial 4.0 Attribution International License

As you can see, I did not always get a red truck. Here is the image I decided to keep.

Used under Creative Commons Noncommercial 4.0 Attribution International License

I was using a free account which allows the generation of 20 images. If you request revisions, those 20 images are used up quickly. After that, you must pay to play. At $8-$10 per month, it is cheaper than most streaming services.

Will I pay to use it? Perhaps. I need to better understand the copyright and licensing. I think it is a great tool to illustrate fictional stories.

Will artists and designers get upset about this? Does it hurt their ability to earn money based on their own talents and hard work? Will their work posted online be used as source images to generate AI images? These are the unknowns. And all very valid questions.

I do know this. AI is here to stay. In some way, we will all need to embrace it as either a user or a consumer.

Please drop me a comment and tell me what you think about this new technology. Have you experimented with it? If not, will you use it in the future? I would love to hear your thoughts.