Yesterday, I released my first product: 'Art Direction with ChatGPT' — giving most of it away for free.
For the past month, I've been working on a framework / method, that aims to solve a specific problem:
Generating consistently styled illustrations, and keeping e.g. character identity consistent through multiple prompts.
I've been testing, refining, and putting a lot of late hours into the project.
Yesterday I finally reached the point where felt it was good enough to release.
I'd love some feedback, and super excited for the hard part of the journey — distribution.
If anyone wants to try out the Style Recipes I've made, I'd be happy to provide a free copy :) Just shoot me an email at hello@fjordkit.com
The idea
ChatGPT is great at generating illustrations, but if you've ever tried to use it for content, e.g. website illustrations, visuals for blog posts, social media etc, you've probably also experienced how much output can drift from prompt to prompt.
I wanted to fix that, so I could generate content that looked intentional and consistent.
And save hours of prompt trial and error.
The product
The outcome of my efforts are two-fold:
A free framework I'm releasing for everyone to use.
A paid extension of the framework in the form of Style Recipes — a set of ChatGPT directives, that consistently makes ChatGPT output illustrations in a certain style.
The free framework also offers a light version of a Hand-drawn Doodle Style Recipe, so you can try it out, modify the Style Recipe yourself if you like — or get one of the pre-made recipes.
Example: Flat Vector Style Recipe
Once you've set up your ChatGPT project, you can simply just upload a Seed Character, and ChatGPT will utilise meta-prompting to generate the illustration.
Here's some examples:
