1. Poor at following instructions. It can take several hours and dozens of prompts to get an image close to that you actually want, and it then has to be blended or photoshopped to make it usable. If the prompt is vague and you don’t mind what’s produced, it’s great.2. The AI cannot count. Ask for three cars, six cats or eight flowers and you may get a hundred cars, one cat and twenty flowers—or a tractor a dinosaur and a bunch of bananas. 3. The AI has no concept of relative sizes. If you ask for a giraffe and a rabbit, the rabbit could be represented larger than the giraffe. 4. It has no idea of relationships, so any locative prepositions (before, after, behind, inside etc) will be ignored. Ask for one person walking in front of another and you may get them reversed or side by side.5. The size of image requested seems irrelevant. Ask for wide and it will simply do as it likes and leave the edges plain or fill them in with a random colour. No amount of requests for ‘wide screen’ or ‘fill the frame’ make a difference.6. For specific images, it is simply not clever enough to understand detailed prompts and so substitutes nonsense, like adding in extraneous backgrounds and random items. It inserts misspelled versions of the prompts onto the image. If asked for a cat and a duck, you may get cats with duckbills, or ducks with claws and whiskers. 7. Changes style as it pleases – photo-realistic, line drawing, black and white, vibrant pen and ink - with no discernible pattern and no means of maintaining a style if several images are required in a similar style, even if the prompt remains the same and only one or two details change.8. The cut-out and erase tools are poor. The cut-out tool is amateur; the erase tool crude, and it is frequently denied when MSD decides it can’t erase the thing you want removed.9. Frequently claims prompts have violated generation guidelines despite their being decent, accurate and utterly innocent. That says more about the way the AI has been trained than it does about the creator writing the prompt. It doesn’t matter how you rephrase (using every known synonym, paraphrase and alternative known to man OR AI) it still come sup with its standard and inaccurate response “Content wasn't generated because your prompt includes wording that may be in violation of Designer code of conduct. Please change your phrasing and try again.” 10. While users are encouraged to provide feedback, it is a pointless time waste To respond to every poor image, mistake, and error would be a full-time job. In addition, feedback is clearly not acted upon as the same problems persist. 11. Any prompts that mention a young woman, a princess (for fairytales) or girls, usually return the kind of images seen in girlie magazines, even when you ask for the image to be suitable for children. The AI has been trained to show women as sexual objects. 12. Trying to contact Microsoft to ask for help is hopeless. Average wait time 2 hours (turned out to be three and twice they dumped my ‘chat’ saying the issue had been resolved when I hadn’t chatted or spoken to anyone). In the end I put in a new request for a callback; by the time they contacted me it was late at night and I missed the call because I’d gone to bed. Extremely poor customer services.If I had an alternative, I would ditch Microsoft tomorrow.
Claim your business profile now and gain access to all features and respond to customer reviews.
