Claude Monet - Water Lillies - Metropolitan Museum of Art, NYC
Hope you are having an amazing week! I’ve been fighting a cold this last week, and I was awake last night trying to calm a coughing fit. And in the middle of it all, I had a thought about a Claude Monet painting that I saw in the MET in NYC in January when I was there. It was this:
“If I was Monet, and I was imagining this amazing painting, how would I describe it to someone else that I wanted to create was I was seeing in my head?”
It made me start thinking about AI, especially when it comes to art.
AI is mostly just guessing. The more you tell it, the better it guesses. But can you really create what you see in your mind?
Art is a Hallucination In Our Minds
As an artist, I tend to see what I want to draw, or paint, or create in my mind before I actually make it. But I need to put one stroke, one line, or one polygon down to get going. Then I can work my way to eventually create what I see in my mind.
There’s lots of great discoveries along the way, but usually, I have a very specific, clear image in my mid that I’m working towards on the paper, canvas, or on the screen. With each addition, I’m getting closer and closer to the final image.
You could call this “hallucination” like an AI diffusion model works, but there is one simple difference: I have a target that I see in my mind that I’m working towards. The hallucination is just the steps to get to a target.
With AI diffusion, it’s generating an image based on the pixels and noise from the previous sample. There’s no template. It’s just trying to put everything you are prompting it with into a cohesive layout. Each sample looks at the previous one and says, “well, that looks like _____, and that’s what they prompted for, so let’s keep going.”
Some may say, “Well, you can give AI a referece image and it can match the reference.”
True.
But is the reference EXACTLY what you see in your mind? I can almost guarantee not. You’ll still be getting a best-guess.
AI is Just Doing What it Knows
Remember, AI is not imagining or thinking on its own (yet). It just does what it knows based on what you give it. So you need to give it the most accurate information you can. You can’t expect it to read your mind (just like you wouldn’t expect that from other people), so you need to provide it with as much as you can, and not count on it to give you amazing results.
And even when you give it information, be ready to see completely different images than what you intended. That’s part of the iterative process, and completely normal. Be patient, but also be willing to iterate with it… a lot.
Why is this Important to Product Brand Companies?
One of the biggest drawbacks to using AI for product visualization is something that is absolutely necessary to make sure your product is accurate.
Control.
If you’re okay telling someone else what you have in mind, and then letting them run with it, then AI might be for you. But if you have a very specific vision in your mind, you want to be able to drive the content yourself, controlling the outcome.
Yes, you could guide your AI creation with other imagery, like 3D screenshots, or drawings, etc. But then you have to create the reference to feed AI. Does it maybe streamline the process? Yes. But does will AI do all the work for you? No.
In other words, don’t think that by using AI, you’ll eliminate the need for human creativity and direction. You’ll speed up parts of the process, but AI still needs to know WHAT you really want, and usually more than just a prompt.
Product visualizations need to be accurate. They need to look exactly like the real thing. They can’t just be approximations. And if you leave it all up to the hallucinations of AI, you won’t have the control you need, and you’ll not get the outcomes that you want.
This is the perfect segway into a discussion about how 3D is needed even more now, but we’ll leave that for another day.
One thing that I love doing is talking in person with individuals at companies who are trying to figure out their 3D/Ai pipelines. If you would like to discuss what you are doing with your product pipeline, and would like to know more about how you can make it more powerful, efficient, faster, and more flexible, set up a quick disocvery call with me. I look forward to meeting with you!
About this newsletter: Every week, I share some insights into 3D content creation pipelines - tools, processes, success stories, pipeline mistakes, and more. It’s meant to help those involved in 3D product creation, from artists to managers and leadership, understand what is happening in their world, and how to use the latest technologies to create higher quality with greater efficiency, saving time and money.
If you have received this newsletter in error or wish to no longer receive it, feel free to unsubscribe. I hate to see you go, but feel free to come back to learn more anytime here.
P.S. If someone forwarded this to you and you'd like to get future issues, you can subscribe here: