OP
- Thread Starter
- #51
The previous post (the one by hern42) deserves to be more noticeable, so read it and like/thank it, everyone.
Crumb 1501: Authorial Latitude: I approve some of the images generated by ChatGPT as graphical representations of my crumbs even when they have clear distortions in some secondary elements, usually in the background, because they lend authenticity to the whole experiment—this isn’t reality; this is all make-believe, and I find it enriching that the dents left by the tools show. Additionally, sometimes I have no idea why ChatGPT included certain elements in the image, but, if they look well rendered, they pass.
Text by Alfredo Llaquet-Alsina.
Design and image by ChatGPT.
MY PRODUCTION PROCESS
Note: If you notice a typo in this image or any other, please leave a comment. Usually, I can fix them.
Crumb 1501: Authorial Latitude: I approve some of the images generated by ChatGPT as graphical representations of my crumbs even when they have clear distortions in some secondary elements, usually in the background, because they lend authenticity to the whole experiment—this isn’t reality; this is all make-believe, and I find it enriching that the dents left by the tools show. Additionally, sometimes I have no idea why ChatGPT included certain elements in the image, but, if they look well rendered, they pass.
Text by Alfredo Llaquet-Alsina.
Design and image by ChatGPT.
MY PRODUCTION PROCESS
- I write my crumbs (very short stories or reflections) when I feel like it.
- I have 9 prompt templates that instruct ChatGPT to convert one of my crumbs into an image using a certain style.
- Each template basically says to ChatGPT: create an image in this particular style that reflects what the crumb says. Add the text of the crumb on the image, plus "Alfredo Llaquet-Alsina presents".
- It's important to note that ChatGPT decides by itself what specific image to generate every time. I just provide the crumb and the style. It provides the rest. It's its "creativity" that decides how the image will be.
- To speed up the process, I use a custom FreeBSD script that gets the crumb from stdio (this means I paste it on the terminal), and consecutively pairs it with each of the prompt templates and copies each result into the clipboard, one at a time, waiting for a key press in between.
- I paste each final prompt into ChatGPT.
- ChatGPT creates the image.
- If the image doesn't look distorted and the text is correctly rendered, it gets approved.
- If the image is okay and I like it, but the rendered text has a few errors, I may decide to fix the text manually by copy-pasting individual letters, which is tedious but worth it if it means saving a nice image.
Note: If you notice a typo in this image or any other, please leave a comment. Usually, I can fix them.