Insert_Head_Here
Its a fun LoRA that replaces, or removes heads, the idea and a small section of the dataset came from @freckledvixon . Overall its kind of easy to use, but a few of the concepts, certain objects, or animals, can take some promptcraft to get working especially when they are heavily trained in the checkpoint.
when using this, rather than saying 1girl or 1boy do your best to avoid using human terms instead opt for female_focus or simply describing clothing, that seems to work best at maintaining the concept.
be sure to change weights if something is too strong in prompt set the weight down a bit like so (thing:.8)
TRIGGERS:
h3ad - the overall tag, sometimes useful to strengthen the model concepts
headless - trained on headless characters all by itself its still sometimes hard to remove the head but this trigger can be super helpful with certain objects that are small or would be floating where the head would normally be (easiest way to get headless is to add smoke because a certain anime character exists)
animal_head - this is for animal heads
object_head -this is for object heads
special note: because of how lora training works there isnt a super solid line between concepts and you might actually get results you like more by using animal for an object or vice versa.
NEGATIVES:
these may not be necessary on all models but they helped me
(human,fur,head,mouth,nose,eyes,hat,holding:1.5) - half of these may not do anything really noticeable I just ended up adding on to my original negative as I saw things that interfered with the concept (DONT KEEP HUMAN IN NEGATIVE IF TRYING TO GET AN ANIMAL unless you want it more furry leaning)
source_furry - to help assist with maintaining a human body
tail - to help avoid having tails if not wanted
add or remove tags based on what you are wanting to do.
EXAMPLES:
h3ad,object_head,chinese noodles ,black hoodie,thighhighs, , score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, food_focus
(human,fur,head,mouth,nose,eyes,hat,holding:1.5),cute,source_furry,plant,flowers

animal_head,crow , animal_focus,black hoodie,thighhighs, score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, side view
(fur,mouth,nose,eyes,hat,holding:1.5)

headless,smoke,black hoodie,thighhighs, score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up,v, peace
(human,fur,head,mouth,nose,eyes,hat,holding:1.5)

Description
FAQ
Comments (16)
EXCITEDDDDD
really cool idea, gj
can you provide a workflow picture? thanks.
if you hit the download button most images should have the workflow still attached i think, but the part of the workflow i used is just a straightforward efficiency node lora stacker>loader>sampler, and for the upscaled ones iterative upscaler
It allows for some abstract stuff and that's pretty rad, great job with this!
"The head holds the questions; the mind holds the answers, but only when we listen closely." - Z
Thank you for the LORA! 💜
Basement Jaxx approves
Cool stuff 👍🏼
Эх! Где была эта штука, когда я делал грибного человека?
Can this work with object LoRAs?
i havent tried, but its possible, when testing i was able to replace the head with some pretty random objects. large objects like cars or houses are what would be very difficult to get working
Is there any way to add additional "training images" to this model so that I can get certain desired outcomes? I'm finding that I'm having to make a BUNCH of images to get like 1 image of exactly what i want...
This is great, absolutely love it!
Illustrious?
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.















