Hey there!
IL recommendations:
Positive: masterpiece, detailed_eyes, high_quality, best_quality, highres, absurdres, 8k
Negative: poorly_detailed, jpeg_artifacts, worst_quality, bad_quality
Pony recommendations:
Positive : score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, masterpiece, best quality, absurdres
I really hope you enjoy the model — I'm super excited to see what you create with it!
If you like what I do, leaving a like, a comment, or some feedback would be very cool — it keeps me inspired to keep creating! 🖤
You can also follow me for more, or if you’d like to support my coffee addiction ☕ (and my creative madness), you can do so here: https://ko-fi.com/jedas
Check out my Instagram here: https://www.instagram.com/jedasaiart/
Stay awesome, stay healthy — and have an amazing day! ✨😊
Description
FAQ
Comments (16)
Does v3IL do 2k resolutions since it's 13GB?
Thats a good question since i dont know how that happened to be honest :D
@Jedas Lol, thanks. Just wondering, since some 13GB models are just unpruned down to ~6GB, but some allow for 2k resolutions (~1536x1536) instead of the usual ~1024x1024 SDXL standard resolutions. Guess it just depends on what size images it was trained on. I'll have to test it out and see what it does.
@Jedas Haha yeah are you going to be able to fix and re release at the normal size?
@jboom91 if i figure out the problem i might do so but if not the next version will be back at normal size for sure !
Wild guess... its 16p rather than 8p (i think it was "p" ; or it is because - as already suggested - un-/pruned).
That said, my "default" resolution/aspect ratio of images that i do is 1536x640 (21:9) - before i do img2img fix&upscale y 1.5, before i do regular upscale by 1.5 (end result ~3440x1440), havent tested this one out yet - and since i only have 8gb vram, probably wont be able to.
EDIT: uh it does work with 8gb vram (+16 regular, so i guess 24? ^^)
@Sephrajin yeah i dont know there seemed to be something wrong in forge i cant figure it out since every checkpoint i used had the normal size and i did the same merging process as i usually do sadly i cant reproduce the checkpoint aswell since i just used wild numbers for merging and i actually didnt wrote them down :D
@Jedas Just an update, but this model seems to handle 2k resolutions just fine. Been testing at 1536x1152 and no weird issues with stretched or double subjects, but for some odd reason, the lighting and background quality seems to improve quite a bit.
can you make a smaller version? thanks
V3IL has beautiful colors, nice contrast, very good realism. I use it with euler a/karras 35 Steps and 4.0 CFG and i like the Pics. Good Checkpoint, thank you for this. And it works very well with DMD2
Just tested all 3 versions side by side and V1 still seems the most realistic. The skin textures and effects on skin seem much better
in case of realism, i think you are right, but in my eyes V3 makes prompt understanding better
Gorean Hmmm maybe, but there are other models far superior at prompt understanding so I tend to use this as hires or img2img. If you want to only use one model then perhaps the later models are easier to work with
JedpointReal v3IL is a fantastic model and has the best realism/IQ out of any current Illustrious model without sacrificing prompt adherence. Also works very well with Pony/SDXL LoRAs, which most Illu models don't. Well worth the 13GB it's using.
Great job Jedas!
woah that looks gorgeous, yet it seems to be corruped a bit (hence a 12gb file size i guess =) - looking forward for V4!
Good!!! Beautiful and accurate generation!
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.