Support my work on Patreon and Ko-Fi and get access to tutorials and exclusive models.
Rate and leave a like if you enjoyed it, and follow for new models!
LoRA trained to generate a person in public surrounded by people, such as photographers. Works with other LoRAs and various models.
UPDATE - new training and small improvements
Updated training: used LyCORIS instead of LoRA, trained on larger dataset with more realistic images, cleaned up tags.
Updated output: less likely to have a camera on character's hands, more detailed crowd, a bit more consistent, you can better increase the crowd or photographers by increasing the trigger tag's weight.
Main tags (use one or both for different effect): public, photographers
// OLD VERSION (1.0) //
Main trigger word: public Other tags used photographer,phone screen,camera,foreground
You can use your usual CFG and weight 1.0
See weight comparison 0.0 vs 1.0, minimal effect on style and composition:
Description
Trained on NAI using kohya GUI
218 images (unedited, not cropped), auto tagged with WD1.4, added "public" token, "keep token = 1" and shuffle tokens.
218 images * 10 repeats * 5 epochs / batch 6
LR 0.0002, AdamW8bit, constant
512,512 resolution, bf16
FAQ
Comments (15)
thank for this concept
I love it!
Good.But how to avoid the camera or other objects on the hands of women?
The training data did not include images or tags related to the subject holding a camera or another object, that has to do with the model you are using which sometimes interprets prompts in that way. This is not an exact science, so you may need to play with the CFG, weight, prompt and seed (or var. seed) to accomplish that. Impainting would also work.
@psoft ok, thank you for your answer
@psoft how do you exactly contained a prompt inside (bracket) so it doesn't influence other (bracket), is it even a feature that i don't know of.
@nicetry20010 I’m not sure I follow, can you provide an example?
What does this promt construction mean?
catwalk \(modeling\),,
\ - these bars
catwalk \(modeling\),, - also 2 commas at the end
the commas at the end are probably just a mistake.
the tag "catwalk (modeling)" is based on danbooru/WD14 tags. the bars are part of the way to prompt when you use parenthesis as part of a tag like "catwalk \(modeling\)", otherwise the parenthesis would be processed as the regular syntax like (yourprompt:1.2)
2.0 plz
Is there any way to set the crowd to be all women? I love this idea but sadly don't want crowds of men :/
@cads300 not really... the training data does not contain anything of that sort, and unfortunately, I haven't been able to source data on that for the new version.
Details
Files
public_v1.0-000005.safetensors
Mirrors
public_v1.0-000005.safetensors
InPublic.safetensors
public_v1.0-000005.safetensors
public_v1.0-000005.safetensors
Public.safetensors
inPublic_v10.safetensors
inPublic_v10.safetensors
inPublic_v10.safetensors
inPublic_v10.safetensors
inPublic_v10.safetensors
public_v1.0-000005.safetensors
inPublic_v10.safetensors
inPublic_v10.safetensors
inPublic_v10.safetensors
inPublic_v10.safetensors
public_v1.0-000005.safetensors
in_public.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.