Been really exciting to experiment with AI art generation.
Text prompts fed into the various systems often produce strange results, that are incredible when ya think about it.
For a digital artist it cannot be ignored that the systems couldn’t do this without scrapping a lot of imagery made by humans uploaded to the web.
It is pretty crappy that companies are charging people to use these tools which have been trained on artists work without their consent or any form of royalties. What makes it worse, is the ability for people to include an artists name in the prompt to further clone the style.
Artists copy/steal from other artists with the eyes to learn and grow but when an algorithm does it, its worse. It’s also weird people say,
“look what I did” when in fact its the algorithm that did the work.
That’s my 2c anyway.
– –
When testing the above 3 AI tools, the best results achieved are when adding a human gallery and render engine as a reference prompt – like:
” artstation, behance, deviantart, octane, redshift, render, digital art. “
Even better results can be achieved when included an artist by name…
If the artist had sold some rights to the AI company or got some sort of dividend it would be a different matter.
Without including the source of the style or reference for the AI – the artwork produced is pretty terrible. Things are moving fast in this space so it is quite concerning. Eventually it will just make good art with enough training without having to reference the artists by name.
The images above are links to pages showing experimentation results.