Reviews
MOSS Image
Overwhelmingly Positive
99% of 1,605 reviews recommend it · 1,605 reviews
1,6023
2 words: HUGE VARIETY. Poses, faces, clothes/armor are different every time, lots to choose from. Excellent for original art, less useful if you want to reproduce a particular character or simply want more consistency.
Decent understanding / prompt adherence.
8-step Turbo LORA is a MUST. Not only it's much faster (cheaper), the results are somehow MUCH better than CFT without it.
0 comments
Read review by MV261
.jpeg)
This model is amazing! I almost skipped it, but it's definitely worth trying out. :)
0 comments
Read review by Silmas
rotechacon667
· SFT
Recommends
Looks promising, Can't wait for the anime version
0 comments
Read review by rotechacon667
The_Doggy_Enjoyer
· SFT
Recommends
Very good model, it creates very looking good images. i found annoying the fact it's very sensitive for training, but i loved the challenge on those 15 days i spent learning about how it works in practical ways (testing and failing a lot). Also loved the fact it's small enough so i can train LoRAs locally.
0 comments
Read review by The_Doggy_EnjoyerMU
mu5ic3oy337
· SFT
Recommends
Excellent for a newly released model.
0 comments
Read review by mu5ic3oy337
Otto_Mattiq
· SFT
Recommends
The text-encoder is definitely the biggest strength of this model. As a test, I fed it some of my more complex Krea2 prompt and, hmm... Prompt adherence is on par, but the visual output is no Krea2. At first glance, MOSS seems to be aimed mostly at realism, with 2D illustration and 3D styles that didn't impress me.
Reminds me a bit of Ernie and even Anima: prompt adherence that highly impressed me at first but visual output that I didn't like enough to stick with the model.
OOBE fully uncensored sounds like a big and interesting bonus, something to test in local gen.
0 comments
Read review by Otto_Mattiq