Image generation on ChatGPT is 1,000x better today than it was last week. How could you use this for your MSP's marketing?
You see that image? It's supposed to be me.
I uploaded my headshot to ChatGPT this morning, told it to place me inside Doctor Who's TARDIS (a lifelong fantasy), and put a speech bubble in my mouth telling you this news.
It replicated the time machine, my shirt and my face wrinkles just fine 🙄 The rest of my face is... weird... uncanny. Clearly not really me.
But, let's look at the bigger picture. This image generation on ChatGPT is 1,000x better than what you could do this time last week.
A couple of days ago OpenAI released new image generation within its current 4o model. You can see the details here.
It can create better images. You can iterate more easily. And for the first time ever it can handle text properly without spelling mistakes and nonsense creeping in.
Here's an example of how much better it is. I fed this prompt into ChatGPT last week under the old model.
Create an image. I want it to be widescreen size. It should be in a fun cartoon style. It should show an IT technician holding a stack of software boxes. All the boxes are productivity software
And this was spat out:
Look closely at the text, and the random elements in the background. What is poftwate? What kind of software is Infox Blastro? 🤣
Then I put the exact same prompt into the latest ChatGPT 4o model this morning. And got this:
Stylistically it's totally different. It's very simple. But it's also more accurate. See how it even matched up the colours of the software boxes to the 365 applications?
I uploaded the last week's image and ask it to match the style. Plus, I was more specific on what I wanted displayed on the boxes.
Do it again. I want you to make 2 changes. 1) Make the style more like the image I'm uploading. And 2) Change the words on the software boxes to say things like "productivity" and "get things done"
Getting there. Although I spotted a spelling mistake in a poster on the wall.
There is a spelling mistake in the poster on the wall (productvite). Correct that mistake and render it again
Wow. The image iteration is a lot better. It's still generating the whole image from scratch rather than changing the incorrect element. But it's way better than last week, when it would have generated a completely different image.
Note: The flip side of this improvement... it's also a LOT slower to generate. I'm talking minutes with some of the tests I've done today. I'm sure it'll speed up over time.
So... what does this mean for your MSP's marketing?
You could use it to generate original images for your website's blog or LinkedIn. It opens up a whole host of ideas.
For example, you could generate a photo real image that illustrates some content with your city's famous landmarks in the background.
Or you could give it a photo of you and your technicians. And use yourselves as characters in your images.
What ideas do you have? Comment and let me know.
PS in my MSP Marketing Edge we use ChatGPT to generate images for communications to our 700+ members, but NOT the images supplied for them to use in their marketing.
Sure we use AI to try out ideas and create elements for within the images. It's a useful production tool. But the images are created, composited and checked by my human team.
Why? Because I believe that right now, trained professional humans still beat AI hands down for professional content creation with 100% control.
PPS this pop-up was on the ChatGPT window today... looks like they're really going after Google's lunch...
