Which direction you are facing today?
If you have any request or would like a commission, feel free to comment or message me here or on discord, I will see what I can do!
Experimental concept LoRa that puts character on left or right side of screen.
Works best on landscape and it's recommended to use it.
Horizontal and square resolutions have a bit trouble but it's workable at higher LoRA weights.
The smaller character is on screen the better LoRA will work
Please avoid using too many LoRAs as it will weaken this LoRA and make it useless
LoRA weight strength : 1-1.3
Main trigger :
Should work without any trigger, but might help.
left-side
or
right-side
wide shot (while it's mostly used for presenting scenery it also helps with putting characters to the side
Secondary triggers :
straight-on or towards viewer (straight-on tend to put character in the middle so lower weight might be needed)
full body - make it easier to put character to the side as it changes perspective and size of character but not necessary
out of frame, crop split - can put character partially out of frame further moving character to specific side, but it doesn't work that well, hit or miss. Mostly miss
Recommended negatives :
Depend on tags you use.
If you are using towards viewer, you can add from side to negatives
SD 1.5 :
Most visible on landscape 768x512 resolution, you can even go for 1024x512 and still receive very good results.
It will work on square or portrait but won't be this much noticeable
But if it will help someone, here :
lside, left side
rside, right side
If you enjoy my work and you would like to support me, how about a coffee?
Description
FAQ
Comments (11)
nice
Is there any possibility of a companion lora for these to get characters looking/facing left or right?
@MisterL If I understand correctly, you would like a character to be standing, for example, on left side but facing right side?
You make hyperspecific LoRAs that nonetheless look extremely useful!
Thank you! ❤️
Yeah, whenever I make concept LoRA, I try to make something that checkpoints don't know or are useful for composition.
Usually idea comes from experience if I can't make something with just prompt
looking for positioning,got this GOLD lora instead
Are you able to turn this into a slider? that'd be incredible! like -3 being left edge of frame, +3 being right edge, and 0 is centered?
I never done slider and from what I know it uses completely different training method. I will have to research and see if I'm able to do it, but no promises at the moment :(
@MLGnom for sure, your LoRA is so crazy useful, helps fight all these centered 1girl portraits and makes stuff interesting :)
This is very interesting, but doesn't regional prompting achieve the same thing?
Yup, but not everyone have access or ability to use regional prompting (on site gen for example). From what I heard, people do use my LoRA together with regional prompting for best result as it makes positioning easier. ^^;
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.









