Update
- I like the original better 5
- I like the new one better 11
- I think the new one is good, but the original is still better 2
- With some changes, the new one could be better than the original one 2
Hi there everyone, its time for another update! o7It's been a while since the last one so let's go through what's been going on.
Since the beginning I've been doing my work using the same model, but recently I've been trying to upgrade to a new one. As easy as this is, the main factor holding me back from doing so was me not being able to reproduce my default style in the new model.So I've been trying, tweaking and all to try to produce something similar to the current one.
Doing so I've come up with a style that I liked very much (might've seen it in one of the recent commissions on discord) but is vastly different from my current one. After some further trial and error I think I'm getting closer to cracking this case.While reproducing my exact default style would be very hard to pull off without training an entire model on doing so, I'm trying hard to try and make this happen.
So without further ado, lets go through how it's all shaping to look like:
I've used the same prompts for every image, just changed the styles and the model of course. You'll see that this model handles some prompts a little differently, resulting in slightly different poses, or most noticeably: expressions.This doesn't mean that every expression will be like this, just that I'll have to prompt them differently to get the same results. But for the sake of consistency I haven't re-prompted them for this test.
The slightly shy, curious expression turned into a more "tsundere-like" one
The lights and shadows are in general handled differently, if its better or worse I'll leave it up to you guys.
More dynamic shadows and reflections give more detail to the image.
Backlights are handled properly.
Yeah she looks a lot hornier in the new one, the "lustful" expression was originally prompted for in the image set but I forgot about it. The old model couldn't really recognize it so it went with a more basic smiling, happy expression. Which might actually be better in this case, in my opinion.
More dynamic poses in general, and I think the characters are better placed on the frame, with less parts sliding off the screen.
And while I can't confirm this a 100% right now, but it looks like the style retention is better on the new one. The default model was actually holding up really good, even with the slight shifts in styles, I've pushed it to the limits with some pretty wacky resources so I'm surprised this could handle them this well.
So this is the rundown for todays update. Please let me know your opinions either in the comments or by contacting me directly here on Patreon, Discord or X (formerly known as Twitter)!Is it better? is it worse? you can be as brutal as you'd like in your criticism I don't take offense 😄I'll do another post with the new model in the near future featuring some actual "action" as well, so keep an eye out for that one!
I've attached the comparison images to the post, so feel free to check them out in full resolution there as well.I'll leave a poll for some easier reviewing if its more comfortable for you this way👇
And as always huge thanks for all the support, and have a wonderful Weekend!