Microsoft Designer’s Generate Image feature has just received a big boost with the upgrade to Dall-E 3 ‘text to image’. That means a significant improvement in the quality of images made for you in Designer.
That means the images made in Designer should now be more realistic, more detailed and text should be rendered better.
Designer should use Dall-E 3 now but you can check from the (i) tooltip.
Dall-E is made by OpenAI who received funding from Microsoft. In exchange, Redmond gets to use the same technology in a modified form in their AI offerings.
Same resolution
‘Realistic’ does not mean bigger or better resolution. Designer generated images are still 1024 x 1024 pixels at 96dpi and 24 bit depth, mostly around 3MB. That should be enough for most Designer graphics.
If you need larger or high resolution images, look at other AI image makers like MidJourney.
The original Dall-E 3 is run by OpenAI (along with ChatGPT) and has features like Variations and Generation Frame (aka Outpainting) that aren’t available in Microsoft’s version of Dall-E.
Microsoft Designer: Straight Talk has a whole chapter explaining how to make images from text in Designer, Dall-E and other AI image makers.
Any Designer images you make are saved in the ‘Generated on Designer’ folder under ‘My Media’ (though this keeps changing, so you may find some under ‘Uploaded to Designer).
Fixing Text in generated images
Dall-E 3 does a better job with text but hardly perfect. In many cases it’ll still mess up the letters. Here’s a result from the request ‘Movie Poster for “Microsoft Designer Straight Talk”’ which should be straight-forward but Dall-E misspells and adds a new letter to the alphabet.
In many cases you’re better off using Designer text blocks. Taking the image above we used Designer’s “Generative Erase’ to remove most of the text (while keeping the background) then added the right text.
All this and a lot more is explained in, you guessed it, Microsoft Designer: Straight Talk.
Dall-E 2 for comparison
For comparison, here’s what Dall-E 2 does for the same text prompt as above. As you can see, the text isn’t even close.
Needs better tagging
Unfortunately, Designer’s AI made images still aren’t properly tagged in the image file Properties. Nor is the original text request/description saved. The result is a lot of images with no way to know what exact text was used to make each one and that’s vital detail customers need to make more images like one made in the past.
There are image fields like ‘Comments’ and ‘Program Name’ that could be used to fill in the prompt and ‘MS Designer-Dall-E 3’ respectively.
Microsoft Designer: Straight Talk
Resizing for Office documents in Designer
Copying images from Designer to Office documents
Designer mobile now on iPhone and Android