I have been playing with just a small number of AI generators. They have been fun and frustrating to use at times since I have typed in the prompts according to the English language but the generated images are not what I have typed in.
One would think that the AI programmers would have integrated the language, grammar and vocabulary into the generators. However, not quite. I would have keyed in the following prompt but the images would have missing elements.
So, the above was just a simple prompt that I thought the AI from Gemini would understand. Apparently, not. I was playing around with the generators because I would want to use the images for something. Of course, as a money-making tool ultimately.
One such use is the ability to create consistent-looking characters and most of the time, the same description for a character will end up looking different each time. This is frustrating for authors who are writing stories for children and using those images for their books.
Recently, Midjourney came out with a feature that allows one to create consistent-looking characters and authors of picture books are happy with it. With the use of AI generators, illustrators are in danger of losing their bread and butter because they won’t be hired by authors who will want to reduce cost in producing their books.
Of course, there is a fight about copyrights. This is a never ending problem because copyrights for the traditional methods have always been a problem and plagiarism is rife too. With the generators, and people posting their generated images proudly, there are fights about who got the image generated first.
Now, with the image generators, there are scammers out there. They claimed that the child’s first art piece is this and claimed sobbing stories about people laughing at the seemingly fantastic art piece. When people don’t realise it is AI generated image, they get conned and then get scammed into phishing sites.
There are unsavoury sites, to put it mildly, that use the generators and generate famous characters in doing unsavoury actions in their fantasy worlds. It can be difficult to differentiate real photos from AI generated ones because of the realism being used by the AI generators.
The popular AI generators are mostly Midjourney (paid version only), Dall-E (paid version only), Stable Diffusion (free and paid), Gemini (free and paid) and Leonardo. They each have their quirks, Gemini to me is the weakest in generating high definition images. They tend to be darker or duller in terms of the colours.
Stable Diffusion would be my preferred output for book illustrations because of the bright colours. I prefer my outputs to have bright colours. Next in line would be Leonardo as it is rather flexible too. Leonardo does have filter moderation for keywords like “baby” and I was not able to generate cute babies doing cute stuff. I would have liked to know how Leonardo would generate babies so that I can probably write books for young kids.
Probably the next step would be using the AI images to sell printed items like T-shirts, mugs, tote bags and many others.
