instead of a logo that is fixed or static ... why not propose a logo that constantly evolves based on continuous A/B testing being fed-back into your AI logo generation model?
You host the generated logo as a live API endpoint that renders the "current" version of the logo on demand
This endpoint takes some input parameters 1. to control which version of the logo gets returned & rendered and 2. to provide context to were the logo is being presented, and tags a unique requestID. After the fact, when user interaction/session is complete -- the KPIs about the success of that particular user session is recorded and matched with the logo requestID
Now on the logo serving side, the algorithm continuously introduces small randomized tweaks to the logo -- subtle changes in geometry / relative sizes and shapes / subtle changes in colors and gradients etc. and keeps track of what was served for which request ID
Over time -- you nudge the logo generation towards what seems to drive better customer engagement and outcomes. Survival of the fittest - automated.
Now that we have this idea in the space of logos -- the same could also be used in the marketing copy used on product pages -- where the text is generated through a GPT-like model and continuously evolves to make the impact better?
Same with web-layouts and CSS styles generated through code-GPT models?
How far back can we integrate this into the product/service stack -- instead of just the presentation layer?
Distant future:
You host a fully-automated tech-news website and hook it up to this "evolution" model where there are constant nudges towards what seems to engage audience better and convert into sales. You take a one-week vacation and come back to find out your site is now a rabid polarizing political disinformation generator and a p0rn generator for metaverse.
You can't possibly know that you don't encroach. Part of the package of being a professional logo designer is knowing the field. They literally know hundreds if not thousands of existing logos. There are books and references dedicated to finding dupes. This is a very important aspect of the logo design and ignoring it is an extremely unwise thing to do.
With your process you are guaranteed to generate a logo that will look like a blatant derivative or a verbatim copy of existing work. Consider what will happen next.
It's not a serious B2B effort. It's just a way to collect some money from individuals who want to slap a graphic on their GitHub project or their lawnmowing service.
fwiw I really like the color schemes on the marketplace. Kind of a modern dark neon vibe. But would be cool to select from a list of "moods" or "themes" for someone who doesn't have an exact color palette in mind, more of a vague idea.
I assume you're referring to the package where you buy 10 logos for $14? I am absolutely willing to work with someone if they aren't happy with what they receive.
That's half the reason I built the marketplace. People can see what they are buying instead of blindly making a purchase.
This is a very new area where law has not caught up. Surely in any creative process (book author, painter) there is a process where the artist selects and omits elements before the final product.
Is not the selection and omission of poorly generated images the "human authorship" [1] necessary to qualify? Or can OP modify AI generated art meaningfully to qualify for copyright?
Additionally the choice of the model is an important choice too. Some AI art collections even make their own derivative models where a human picks extra images to train off of.
It reminds me of 3D art where there is a lot of up front work in making models, textures, armatures, and shaders. You send a ton of data to the GPU and it spits out a rendered image or more likely a tile of one. Sending a vector of bones does not seem much different then sending a vector of a prompt.
