Ai-powered greeting cards that wow...

The ability to generate video and images has advanced significantly in the AI world in the last 18 months; you can't deny it.

As a result, we're witnessing yet another explosion of innovative applications, from virtual influencers and youtubers to immersive experiences, that were previously unimaginable.

The boundaries between reality and fantasy are blurring, and the potential for AI-generated content to transform the entertainment landscape is vast.

But before we look at what’s possible now, I was going through my Canva Account trying to find some old banners and stumbled across these early images generated by DALL-E 2:

In case you are wondering, the prompts on the bottom two images were mostly likely not what they perhaps look like!

I was probably trying to create a robot cat or, from vague memory, maybe even recreate a cyberpunk version of Kiwi Claire’s real estate famous cat Fat Basil.

For the image on the left, I thought I would ask ChatGPT to tell me what the prompt might have been, and this was the response:

"A futuristic robot sitting next to a friendly orange and white cat on a grassy field, with the robot gently touching the cat. The cat is looking surprised, and the scene is brightly lit outdoors."

For all I knew about prompting back then, it was more likely to be “A futuristic robot sitting next to a ginger cat” (the end).

I asked ChatGPT to recreate this image today, and here’s what I got:

So I think you’ll agree from the above two images we have come a long way in the last 12 months.

And it will only get better - NB, we haven’t yet talked about Flux.1. Midjourney’s updates and plenty of others - (but to finish this blog post tonight, I may have to deal with those another day.)

One thing that needed improvement has always been text placement

We’ve had several posts on here about how to fix the wonky text that AI generates.

The best way I have found is with a Canva workaround - and you can find a tutorial on how to do that here.

However, generating good text still remains slightly tricky: to generate something really awesome, the model you’re using needs to understand context, semantics, and visual balance, just to name a few of the things it needs to get it right.

However, recent breakthroughs have enabled AI models to generate visually appealing text where the overlays work pretty well with the surrounding image.

So onto the bit you can actually use to impress your clients (or your employees, or partner….) drumroll please…

Tool of the day: Wishesly

I actually had a fair bit of fun with this tool this afternoon; It’s called Wishesly, and it generates personalised AI greeting cards…

I almost certainly won Wife of the Year by popping this unexpectedly in the Slack Channel…

As you can see, the text (and Deadpool) are both pretty realistic and, dare I say meme-worthy!

There are all sorts of pre-sets in Wishesly for Birthdays, Beaches, Graffiti and more and can be sized for Facebook, Instagram and Linkedin:

If you ignore some of the pretty sexist images (sorry had to say that!), there are some cool ideas for personalising emails or creating some cool social media posts.

It’s a paid tool, but not too exxy, $9 USD per month will get you 20 generations.

I should say it’s good - but not perfect…

I plan on trying it tomorrow for some cold outreach emails for the AI-Powered Agents course. I’ll be sure to let you know how it goes…

Happy Hunting 🚀

PS I can’t leave tonight without letting you know that Midjourney’s web interface is now available to everyone. So, if you found that whole Discord thing a headache, this should be good news, here’s how to use it.