Did you create all three pairs at once or did you put them together later in an external graphics program?

parszuki said:
Did you create all three pairs at once or did you put them together later in an external graphics program?

In this instance, only the middle one was generated. It was then re-generated a few more times with 50% variant seeds to slightly alter the limbs and angle, and those copies were placed on either side, then the AI stitched it all together.

terraraptor said:
In this instance, only the middle one was generated. It was then re-generated a few more times with 50% variant seeds to slightly alter the limbs and angle, and those copies were placed on either side, then the AI stitched it all together.

Will you reveal two little secrets? What image generation tool do you use, and what propmt and settings did you use to create this image?
(actually, as you wrote - to the middle one)

I'm interested in experiments/variants using an anthropomorphic wolf instead of a human. Sometimes I managed to change keywords (e.g. cat -> wolf, rad -> deer), add a few of my own and the results were satisfactory.

parszuki said:
Will you reveal two little secrets? What image generation tool do you use, and what propmt and settings did you use to create this image?
(actually, as you wrote - to the middle one)

I'm interested in experiments/variants using an anthropomorphic wolf instead of a human. Sometimes I managed to change keywords (e.g. cat -> wolf, rad -> deer), add a few of my own and the results were satisfactory.

I'm just using regular ol' A1111. This image was from the same batch. You should get something sorta useful once out of every ten or so, so batch accordingly: https://cdn.discordapp.com/attachments/1019203436429987891/1149376537506881668/00015-2351558532.png

terraraptor said:
I'm just using regular ol' A1111. This image was from the same batch. You should get something sorta useful once out of every ten or so, so batch accordingly: https://cdn.discordapp.com/attachments/1019203436429987891/1149376537506881668/00015-2351558532.png

It turns out that I will have to learn more about this issue:/ Currently, my capabilities do not go beyond generating using prompt.
And you are talking about "img2img", i.e. using existing images and further processing them. I don't know any websites that would work in "online" mode, and installing these extensions (for example A1111, StabilityMatrix) is still beyond my capabilities. (Θ︹Θ)