I thought AI image generation would be quick. I was wrong.

I’ve already introduced you to Colin from our Streamline+ portfolio piece. The other character in the module is Meera - the learner’s leader.

Meera pops up throughout the module to offer coaching when the learner needs it, so I thought I might want her in a few different poses and emotional states.

When I started experimenting with illustration styles I went straight to Sora.

Now, I'm sure Sora is very good, so it might just have been my prompting … but some of the things it produced literally made me laugh out loud!

I'm not criticising Sora here at all - this is more about the reality of using AI for image creation. It can feel quick and easy … especially if you want a one-off image.

But if you need a series of images, or you have something very specific in mind, you can end up spending a veeery long time tweaking, regenerating, and starting again.

Two things finally tipped me over the edge, and moved me away from using Sora:

1. Sora really didn’t want to give me Meera on a transparent background.

2. When I asked for Meera to look angry, she changed into a completely different person! (I almost spat out my water at this one!)

Then my kids then got involved and asked me to turn Meera into a video.

My mistake - I didn’t realise Sora couldn’t animate illustrations. When I asked it to make Meera wave, it gave me a video of the image not moving, followed by a split second of a “real” person painting.

Honestly … you’ll see it in the PDF, maybe you can work out what’s going on 😂

In the end, I switched to ChatGPT, which gave me far more consistency and control for what I needed to build.

Lesson learned: AI is brilliant, but it’s not always the shortcut we expect - especially when consistency matters.

I can’t be the only one though.

What are your best (or worst) AI image fails? I’d love a good laugh!

Nicola

Meera Sora LinkedIn Carousel.pdf

Download PDF • 5.64MB