Inspired by René Gruau's fashion illustrations, this LoRA tries to produce images that imitate fashion illustrations.
The training trigger "regr1 sketch/illustration" is probably not required, but it does not hurt to put it in. But you do need to put in "sketch" or "illustration".
Known problems:
The LoRA sometimes produces people with three legs or even missing arm when they are in profile, if that happens, just change seed or try to lower the weight.
The Shōjo version can only produce androgynous looking men, but then that's Shōjo for you 😅
You can try to put in the optional "background is stark white" to see if you get a more interesting image.
Please post your first image directly here (click on "Add Post" above the gallery), and I will give you 10 buzz for the post (If I forgot, just complain loudly 😎)
Description
Mixing in some Niji style images to the training set to make the LoRa generate Shōjo manga style fashion illustrations.
This version can only produce androgynous looking men, but then that's Shōjo for you 😅
This is epoch 10, the rest can be found here: tensor. art/models/808031774128156510/Rene-Gruau-nijiV3-NoCD6A3Cos6-2024-12-16-01:42:50-Ep-10
FLUX.1 - dev-fp8
Trigger: regr1 illustration
Repeat: 20 Epoch: 10 20 512x512 images 4000 steps
Unet LR: 0.0006 Scheduler: cosine Optimizer: AdamW
Network Dim: 6 Alpha: 3
Epoch Loss
1 0.359
2 0.337
3 0.336
4 0.319
5 0.303
6 0.294
7 0.277
8 0.258
9 0.255 <- lowest
10 0.262
FAQ
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.




