7th August: We're looking for some help keeping e6ai.net running, if you have some experience with coding and would like to help please have a look here
Welcome to e6AI! Here you can find a place to explore all the furry creations made by, or with the help of, AI. With AI art generation being such a new and exciting technology we decided it would be worthwhile to have a central place where people can share their various creations, as well as more easily provide additional information about the creation, like what prompts and models were utilized.
For a quick start on how the site works have a look at our Search Cheatsheet, our Rules, and our Uploading Guidelines.
parszuki
MemberDo I see correctly that most of them were created on the same prompt with minor modifications? You went straight to my heart with this series - the biggest beauties with hooves - including the underrated blackbuck. You could try variations of other species, such as kudu, oryx, sambar, or reindeer.
Will you share the prompt data and settings used, or is it a company secret? :-)
Lungfish1223
MemberAll my generations are based off a very similar base prompt with slight modifications made on the spot to account for how well the model handles a particular species, yeah. The ones here are using a slightly outdated prompt - nowadays I also started including LORAs and negative embeddings. As for other species - I have several dozen still in the backlog to upload (including an Oryx and a Reindeer) but I gotta wait until the 10 I posted are approved before I can upload more.
As for the settings - the model is EasyFluffV10.1, the sampler is DDIM, 30 sampling steps (nowadays, the ones here might have still used 45 - it doesn't seem to cause any difference in quality) - but the most important aspect is that they're all img2img from photo references at 0.4 denoise. Many of my generations, including the blackbuck and bison I already uploaded, would not have worked at all via pure txt2img - I tried, there were no results that even remotely looked like the animal in question. The more common species (from the ones I posted this applies to deer and the gazelle, but it also has been the case with most carnivores) do give decent results via txt2img, but the proportions or anatomy usually come out wrong, so I stick to img2img.
I am assuming that it'd be possible to generate enough art from img2img to train a LORA that'd allow for these species to be properly generated via txt2img, but I have zero experience in LORA training, so it's unfortunately out of my capabilities.
parszuki
MemberI'm a beginner and currently I only understand txt2img (additionally, it is not generated locally, on the computer:( - only via the website civitai.com)
I've seen supporting loras like:
https://civitai.com/models/107913/or-orprocapra-picticaudata-or-tibetan-gazelle
https://civitai.com/models/106636?modelVersionId=114528
https://civitai.com/models/106697/or-or-or-equus-kiang
they are great when it comes to the realism of the animal, but they are average when it comes to "anatomical details that are particularly interesting to us":)
Since you can't share text prompts, the only solution is to keep track of your work. Do you publish elsewhere?
Lungfish1223
MemberI've only been posting my work on the Furry Diffusion discord before e6ai. In the future I'll likely try to stay on top of regular posting in both places, but the upload limit looks like it's going to be challenging to keep ahead of.
parszuki
MemberI visit this server too:) I even post stuff sometimes. (cześć ziomek)
Updated
Lungfish1223
MemberI'm a bit more active there since I've been posting there prior to making an e6ai account, but some of my improved generations have never been posted on the discord - like most of those that are already uploaded here. Discord only had the old crude versions of those gens.
Updated
Login to respond »