Site Header

Deviation Actions
Let's talk Bing Image Creator... for those who use it, you may have noticed that you only get one image (not a max of 4) per iteration and it takes a lot longer...
There's a reason for that. Microsoft (aka Bing) has made Chat-GPT4 the default AI rendering image for Bing Image Creator:
In the past, the default (and only) option was DALL-E3. You can select DALL-E 3 if you want:
Some may ask... should I bother? Granted GPT-4 is slower and gives you only one image but presumably it is a better product?
At this point, I will say your results may vary. Or maybe my results. Since Microsoft is optimizing future content for mobile devices first (grumble, grumble), people like @QueenoftheDivas saw the option before it found its way to desktop users. She reported what I had suspected. Slow... but worth it.
My art tends to be very niche (so @Thekobs says and being brilliant, she would know) so I figured a good test for me would be Blenke (the goblin)... since she is relevant to the page of the same name. Tried it with two different Blenke prompts. Here is the Chat-GPT4 result:
And here is the same prompt fed into DALL-E 3:
I got four results but this was the first one so using that...
Now, opinions will vary, but for me, the DALL-E image is far superior on so many levels. The first time I used the GPT4 result, I got a similar situation. In my mind - for what I do - GPT4 is essentially unusable. At least for the way that I prompt.
I suspect most Bing users are aware of the GPT-4 vs. DALL-E 3 option but if you weren't and if you had a similar experience, you can select DALL-E 3.
Note: Once you close your browser, the default reverts back to GPT-4 so you have to change it to DALL-E 3 if that is what you want.
SCAMMER ADVISORY 22
The Magic Inside... a Lego Challenge...?
SCAMMER ADVISORY 21
DALL-E 3: The Last Hurrah
Chatgpt is really troublesome...
They added it as a model in Leonardo, and at least for my style, the results were horrible...
And the worst part is that it's one of the models that consumes the most tokens to run...





