moneycanwork4you cross-posted this post in Project HOPE last year


Text In AI-Generated Images Just Got Better

Next step in generative AI


Image created with DeepFloyd IF

If you've been following the generative AI space, you have probably seen a lot of weird hands and a lot of text that makes no sense. But those days might end soon.

With Midjourney's V5 we have seen a lot of improvement in the hands department.

Prompt: A perfect human hand

And now it looks like the days of gibberish text in images are about to come to an end too. New models that are better at text generation in images have been released and you can try these yourself.

Here’s an example of how the current Midjourney version handles text vs a new model IF by Deepfloyd.

Prompt: An anthropomorphic dog holding a sing "Your text here"

Keep in mind what you are seeing above is just the beginning of a new phase. If one model can solve text in image generations others will follow and the overall abilities available to us will improve.

Let's look at some more examples of IF doing its magic.


Prompt#1: A photo of a "DeepFloyd" graffiti in a metro station, graffiti spelling "DeepFloyd" Prompt#2: A red hat with text "I Love AI" on it Isn’t this amazing? Not so long ago generating an image of a decent-looking human face was hard to imagine. And now we are entering a point where we can begin to customize the text we see in our generations. You have to remember that these capabilities will only get better with time. Want to try this yourself? The easiest way to access this model is through HuggingFace. Here’s a link for you, give it a go yourself and let me know what you think: DeepFloyd IF Excited to learn more about Artificial Intelligence and the mindblowing tools this technology brings to the table? Check out this listof articles. And Subscribe to get future stories delivered to your e-mail.