Whenever i try human photos, there is some wierd manioulation of fingures, faces, perspective etc which make the picture looks little odd. I hope the team is still working on fine-tuning it. I know it is to avoid copyright/content authenticity issue but the images are not looking great due to that.
Tracy.H • Adobe Employee, Mar 23, 2023Mar 23, 2023
The quality of our models will improve with time, and generated content will better respect the prompt. (Merging all similar threads here so the team can access all requests on this issue)
Hi! I suggest adding of control points to determine for example; favored distance between chin and chest, this will make us avoid too tall or short necks.
This can be used in general for human body's proportions for more cute and accurate results.
I wish i could attach an illustrated image but, I do not have permission to upload images.
Merging all similar threads here so the team can access all requests on this issue.
I found a bug on this platform... it is not able to add humanlike faces, i typed girlfriend but it is not creating an actual human like image, hope you guys will fix it... thank you so much
Merging all similar threads here so the team can access all requests on this issue.
@Giddy_personality5D25, already you can see that Firefly Beta is doing much better with faces since your March 23 post. Here are three portraits I made today as a test for this thread.
@IGG Singapore, I think Firefly is improving, but some faces are still not coming out too well. I tried your prompt in Photo mode with a slightly different way of putting it, and came up with these results. Zooming in, the faces are not all the way there yet. Don't give up on it yet!
IMPROVE YOUR AI, ITS NOT PERFECT AND NOT GENRATION IMAGES AS EXPECTED. EVEN HUMAN FACES ARE NOT CLEARED, IN MANY HUMAN IMAGES THERE IS 6 TO 9 FINGERS OF A SINGLE HAND.
Merging all similar threads here so the team can access all requests on this issue.
First of all... well done to Adobe! A great start.
Background:
I'm in the Educational Publishing space in South Africa and finding local, traditional stock is notoriously difficult. So I thought to generate a test image with the cultural attributes which come up in every artbrief.
I like the setting and the overall look-and-feel of the image that was generated.
Generation term: "South African Xhosa woman cooking on an open fire in the outdoors"
Some critique...
At first glance the images are fine until you look at the proportions of the arms vs the body of the subject. This will have to be refined in future iterations of the software, but as is, is a fantastic visualisation tool.
I attach the four options the AI presented for anyone interested.
Merging all similar threads here so the team can access all requests on this issue.
Agree with the consensus that the face generation for either photo or illustration are very distorted and no usable. Am keen to use this for generating images from text to image but right now that is not possible. Everything looks so odd and distorted, even when a keyword is 'lifelike' or 'realistic', no joy yet.
Habe jetzt, zwei drei Bilderstellung auf Basis von Text erstellen lassen und die Gesichter werden nicht immer optimal dargestellt. Sehr oft sind diese matschig.
81
Replies
AdChoices
