the LoRa was mainly trained with these tags:
guillotine execution
gu1llot1ne
guillotine
guillotinefull
scaffolding
basket
lunette closed
lunette open
tilt bascule
head in lunette
blade cuts neck
blade falls
blade has fallen
blade is up
severed head
head falls
head in basket
head lies
head on pole
headless victim
holds head
victim
victim faces guillotine
victim on bascule
victim strapped to bascule
arms behind back
bound wrists
puts neck in lunette
executioner
executioner grabs victim
executioner operates guillotine
executioner straps victim
blood
blood burst
bloody bascule
bloody basket
bloody blade
bloody lunette
from above
from behind
from below
from below face
from side
front view
povbasket
povlunette
A few notes:
guillotine execution
gu1llot1ne
Are main tags, I recommend to use at least "gu1llot1ne". additionally using "guillotine execution" can enhance or change the result.
guillotine - for all scenes in which a guillotine appears
guillotinefull - was trained with scenes in which a guillotine can be seen in its entirety.
victim - for every scene in which a victim appears. It is often possible to do without if you specify a specific person or gender. Can also be used as a negative prompt to exclude victims.