Microsoft has just added the DALL-E text-to-image generator in Microsoft Edge. You can find it in the Image Creator in the Edge sidebar, which then can be opened in a new window. Now, everyone with access to Edge with the Image Creator can immediately create their own images!
Microsoft image creator via dall-e V. MIDJOURNEY
I decided to run substantially the same prompt that I used in Midjourney yesterday, but in Microsoft Image Creator, to compare the quality of the images produced.
1. EYES IN FOCUS: MICROSOFT/DALL-E
Microsoft’s Image Creator/DALL-E performed more consistently in rendering the eyes in sharp focus. As I noted yesterday, I had trouble getting Midjourney to do so. Some of the eyes in Microsoft Image Creator, however, were oddly shaped, but most were round.
2. MORE REALISTIC HUMAN: SLIGHT EDGE TO MIDJOURNEY
I give a slight edge to Midjourney in creating “humans” that appear more realistic (and not fake) on a consistent basis. Microsoft Image Creator tended to create some images that were just off, such as non-round eyes. For example:
Midjourney’s human renditions tend to avoid such irregularities, but sometimes have almost a too perfect, artificial look. When Microsoft Image Creator was on, I’d probably give it a slight edge in looking more like a real human. But Midjourney’s human renditions are more consistent, with nothing drastically off.
3. DEPTH OF FIELD PROMPT: MICROSOFT/DALL-E
I give the clear edge to Microsoft Image Creator in following the prompt for depth of field, which I used as 2.8f. I’m not sure if my specification of a Hasselblad camera with 80mm lens actually did anything. But the images from Microsoft Image Creator do have the cool depth of field you’d want out of a camera like the one below in which the nose is blurry, but the eyes are razor sharp. (I didn’t experiment with other aperture settings on Microsoft, but plan to do so.)
4. USER-INTERFACE AND FUNCTIONS: MIDJOURNEY
I need to play around with Microsoft Image Creator more. Right now, I prefer the functions on Midjourney (including variations and upscaling, downloading the grid of 4 images, and social network on Discord in which you can see the prompts of others).
I think Microsoft Image Creator can render images that look more like photographs, despite some bigger misses than Midjourney. However, the functionality on Midjourney makes it easier to more quickly refine one’s skill in creating images on the Discord platform and learn from others. I expect both Microsoft Image Creator and Midjourney to keep on getting better and more realistic in rendering humans. So, I’d say it’s a tossup on which is better. They both are amazing.