Hello everyone. Welcome back to the discussion about the time I mixed a lot of LoRA. If you look at last week's post, the picture at the end shows that it has something to do with what we're talking about this time.
Yep, that's right. Orange. The results this time all had the word orange (theme) in the prompt. But because orange can mean orange fruit as well as orange color, in some cases the resulting images show orange fruit with inappropriate placement.
Accessories from Orange Fruit
From this first batch, I got a picture where the character is holding an orange peel (?) that has been transformed into a wing shape. And the other 2 images also show their own uniqueness. One has an orange background and the other has a character wearing an orange mushroom (?) hat.
 |  |  |
|---|
![prompt]
# Prompt 1girl, upper body, (cowboy shot:.8), orange \(theme\), <lora:flat2:-.5:lbw=ONL> <lora:artbytokiame:-.6:lbw=1,0,0,0,1,1,1,0,0,0,1,1,0,1,0,0,0> <lora:Houtengeki:-1:lbw=ALL> <lora:gyokai:.6:lbw=1,0,0,0,-1,-1,-1,0,0,-1,1,1,-1,1,0,0,0> <lora:majoneko-08:-.7:lbw=1,0,0,0,0,0,1,0,0,1,1,0,0,0,0,0,0> <lora:llsoccer-08:-.7:lbw=1,0,0,0,0,0,1,0,0,0,1,1,0,0,0,0,0> <lora:maimans-08:-.7:lbw=1,0,0,0,1,1,1,0,0,0,1,0,0,1,0,0,0> <lora:warboob-lora-10:-.5:lbw=1,0,0,0,1,1,1,0,0,-1,1,0,-1,0,0,0,0> <lora:banana-06:.5:lbw=1,0,0,0,0,0,1,0,0,-1,1,0,-1,1,0,0,0>, absurdres # Negative Prompt ((normal quality, low quality, worst quality):1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), (low quality, worst quality:1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), lowres
In the second batch, we have the main image, where the character is wearing an orange-shaped hair accessory (with leaves). Honestly, I really like the result, but of course the hand can't be perfect.
The other 3 pictures don't have oranges in them.
 |  |  |
|---|
! [Prompt]
# Prompt 1girl, upper body, looking at viewer, ((orange \(theme\))), <lora:flat2:-1:lbw=ONL> <lora:artbytokiame:-.8:lbw=OUTD> <lora:Houtengeki:-.5:lbw=ALL><lora:gyokai:.6:lbw=1,0,0,0,0,0,1,0,0,-1,1,1,-1,-1,0,0,0> <lora:majoneko-08:-.7:lbw=1,0,0,0,0,0,1,0,0,1,1,1,0,1,0,0,0> <lora:llsoccer-08:-.7:lbw=1,0,0,0,0,0,1,0,0,1,1,1,0,1,0,0,0> <lora:maimans-08:-.4:lbw=1,0,0,0,1,1,1,0,0,-2,-1,1,0,1,0,0,0> <lora:fairi-us-08:-.4:lbw=1,0,0,0,1,1,1,0,0,-2,-1,1,0,1,0,0,0> <lora:hike-vanitas-lora-10:0:lbw=ON> <lora:warboob-lora-10:-.7:lbw=ONL> <lora:DOAXVV-10:0:lbw=1,0,0,0,0,0,1,0,0,0,.5,1,1,1,0,0,0> <lora:3D Inspired:0:lbw=1,0,0,0,0,0,1,0,0,0,.5,1,1,1,0,0,0>, absurdres # Negative Prompt ((normal quality, low quality, worst quality):1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), (low quality, worst quality:1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), lowres
Alien or Monster?
In the third batch, the result was like this. This is definitely due to the value of the LoRA being too high (or many LoRAs having high values).
 |  |  |
|---|
! [Prompt]
# Prompt 1girl, upper body, looking at viewer, ((orange \(theme\))), <lora:flat2:-1:lbw=ONL> <lora:artbytokiame:-.8:lbw=ONL> <lora:Houtengeki:-.5:lbw=ALL><lora:gyokai:.6:lbw=OUTD> <lora:majoneko-08:-.7:lbw=1,0,0,0,0,0,1,0,0,1,1,1,0,1,0,0,0> <lora:llsoccer-08:-.7:lbw=OL> <lora:maimans-08:-.4:lbw=1,0,0,0,1,1,1,0,0,-2,-1,1,0,1,0,0,0> <lora:fairi-us-08:-.4:lbw=1,0,0,0,1,1,1,0,0,-2,-1,1,0,1,0,0,0> <lora:hike-vanitas-lora-10:-.5:lbw=ON> <lora:warboob-lora-10:0:lbw=ONL> <lora:DOAXVV-10:0:lbw=1,0,0,0,0,0,1,0,0,0,.5,1,1,1,0,0,0> <lora:3D Inspired:0:lbw=1,0,0,0,0,0,1,0,0,0,.5,1,1,1,0,0,0>, absurdres # Negative Prompt ((normal quality, low quality, worst quality):1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), (low quality, worst quality:1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), lowres
In the fourth batch, for some reason the main image reminds me of a ghost from Japan (do you guys know its name?), but the difference is that the mouth (or cheek?) is fused with an orange (or orange ball?).
Meanwhile, the other pictures are still pretty normal, although they have a slight oddity in the fingers.
 |  |  |
|---|
! [Prompt]
# Prompt 1girl, upper body, orange \(theme\), <lora:skijocek-10:-.6:lbw=ON> <lora:banana-10:.4:lbw=INOUT> <lora:gyokai:.7:lbw=1,0,0,0,0,0,0,0,0,0,-1,1,-1,1,0,0,0> <lora:fairi-us-08:-.7:lbw=1,0,0,0,1,1,0,0,0,-1,0,0,0,1,0,0,0> <lora:llsoccer-08:-.8:lbw=OL> <lora:flat2:-.5:lbw=INOUT> <lora:artbytokiame:-.8:lbw=OUTD> <lora:Houtengeki:-.9:lbw=ALL> <lora:majoneko-08:-.7:lbw=1,0,0,0,0,0,1,0,0,1,1,0,0,0,0,0,0> <lora:maimans-08:-.3:lbw=1,0,0,0,1,1,1,0,0,-1,1,0,0,1,0,0,0> <lora:hike-vanitas-lora-10:0:lbw=ALL> <lora:warboob-lora-10:0:lbw=ONL> <lora:yuzusoft-chibiV2:0:lbw=INOUT>, absurdres # Negative Prompt ((normal quality, low quality, worst quality):1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), (low quality, worst quality:1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), lowres
In the fifth batch, the main image has a rather strange and confusing body shape. It's like there are two bodies, but one of them has no head. If I had known where the mistake was, this would have been one of the better drawings.
 |  |  |
|---|
! [Prompt]
# Prompt 1girl, upper body, (cowboy shot:0.8), orange \(theme\), <lora:flat2:-.5:lbw=ONL> <lora:artbytokiame:-.6:lbw=1,0,0,0,1,1,1,0,0,0,1,1,0,1,0,0,0> <lora:Houtengeki:-1:lbw=ALL> <lora:gyokai:.6:lbw=1,0,0,0,-1,-1,-1,0,0,-1,1,1,-1,1,0,0,0> <lora:majoneko-08:-.7:lbw=1,0,0,0,0,0,1,0,0,1,1,0,0,0,0,0,0> <lora:llsoccer-08:-.7:lbw=1,0,0,0,0,0,0,0,0,0,1,1,0,1,0,0,0> <lora:warboob-lora-10:-.5:lbw=1,0,0,0,1,1,1,0,0,-1,1,0,-1,0,0,0,0> <lora:banana-06:.5:lbw=1,0,0,0,0,0,1,0,0,-1,1,0,-1,1,0,0,0>, absurdres # Negative Prompt ((normal quality, low quality, worst quality):1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), (low quality, worst quality:1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), lowres
Grinning / Smiling without Prompt
If you look back at the fifth batch, the main image is grinning, even though there is no grin or grinning in the prompt. And it happened again in the sixth batch (as well as some subsequent batches that I can't point out).
The real culprit was one of the LoRAs. But even though I searched for it by lowering the value of each LoRA to 0 in each batch, the cause could not be found.
I finally decided to remove all LoRAs and use the sample prompt from the batch before this happened.
 |  |
|---|---|
 |  |
! [Prompt]
# Prompt 1girl, upper body, (cowboy shot:0.8), orange \(theme\), <lora:flat2:-.5:lbw=ONL> <lora:artbytokiame:-.9:lbw=1,0,0,0,0,0,1,0,0,0,1,1,0,1,0,0,0> <lora:Houtengeki:-1:lbw=ALL> <lora:gyokai:.6:lbw=1,0,0,0,-1,-1,-1,0,0,-1,1,1,-1,1,0,0,0> <lora:majoneko-08:-.7:lbw=1,0,0,0,0,0,1,0,0,0,1,0,0,1,0,0,0> <lora:llsoccer-08:-.7:lbw=1,0,0,0,0,0,-1,0,0,-1,1,1,-1,0,0,0,0> <lora:warboob-lora-10:-.5:lbw=1,0,0,0,1,1,1,0,0,-1,1,0,-1,0,0,0,0> <lora:banana-06:.5:lbw=1,0,0,0,0,0,1,0,0,-1,1,0,-1,1,0,0,0>, absurdres # Negative Prompt ((normal quality, low quality, worst quality):1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), (low quality, worst quality:1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit, fewer digits, (extra arms:1.2), lowres
So those were some of the events I experienced over the past week while using Stable Diffusion and playing with LoRA. Next week I'll share a similar story but with different prompt or LoRA.
Thank you for visiting and reading my post this time. See you in the next post...
All images and dividers were created by me using Stable Diffusion and various LoRAs.