馃幁 Emotion Puppeteer 馃幁
UPDATE: Now available as an SD 1.5 LoRA and SDXL LoRA
Photoreal faces with exact control of facial expressions
If you like the model, please post your generations!
Safetensors and inpaint models available on Huggingface, along with training details
These instruction apply to v2 only. There's no good reason to use v0.1.
Import Instructions
To get what you want, you should read this.
Use this prompt format for best results
"{A} face, {B} eyes, {C} mouth, {D}, {anything else}"
e.g. "plain face, seething eyes, screaming mouth, closeup portrait, looking away, old man with white hair"
CFG = 5 is recommended
A - Face keywords
plain, unattractive face - my subjective opinion - biased towards extreme emotions
cute, attractive face - my subjective opinion - biased towards pleasant emotions
adorable, most attractive face - my subjective opinion - strongly biased towards pleasant emotions
B - Eye keywords
seething, >: shaped brows - scrunched brows
grieving, <: shaped brows - scrunched brows
widening, eyelids raised very high - brows arched
glaring, lowered brows - eyelids lowered
pleasing, neutral brows - eyes crecent shaped and smile crinkled
piercing, neutral brows - eyelids neutral
squinting, lowered brows - eyelids squeezed closed or almost closed
winking, right eye open, left eye closes
C - Mouth keywords
breathing, mouth relaxed and slightly open
resting, neutral or slight relaxed frown - lips together
smiling, smiling with lips parted - teeth together or slightly apart
screaming, mouth wide open - lower-teeth visible or chin scrunched
sneering, nose scrunched - left upper-lip raised - upper-teeth not bared
singing, mouth wide open and oval-shaped
delighting, smiling with mouth wide open
kissing, puckered lips
gnashing, nose scrunched - upper-lip evenly raised - teeth together - lower-teeth bared
gasping, mouth open and frowning-bean-shaped - chin relaxed
relaxing, slight relaxed smile - lips together
frowning, chin and or nose scrunched - lips together
snarling, nose scrunched - upper-lip evenly raised - lower-teeth not bared
pouting, slight frown - lower-lip extended
grimacing, lips apart and stretched wide - lips apart - teeth may be visible
D - Other keywords
extreme closeup portrait, cropped at neck or closer
closeup portrait, cropped at shoulders
looking straight, head facing forward and eyes looking into the camera
lookway away, head turned to three-quarters, regardless of eye direction
Description
FAQ
Comments (19)
It would be nice to see this as a Inpaint model
Or a hypernetwork so it can be used on top of other models.
@Trahloc聽I never worked with a hypernetwork before, That's why I didn't suggest it, Guess I should learn about this, :)
Meanwhile you should be able to merge with other models and retain these keywords to some degree
@terrariyum聽I wish there was a way to merge where only the differences would be added so you'd end up with a bigger model. I always feel like I have to throw away something potentially valuable when I've got plenty of vram available.
@Trahloc聽seems to me that is what the merge difference is in A1111
Well, expressions can be pretty subtle, so img2img could probably handle it, right?
@kinomi聽I guess soo, Specially with the correct settings and prompts
@Crazyconcepts聽If you start with a 7 gig model and merge it with a 2 gig model you end up with a model exactly the same size as the 7 gig model not 8 or whatever from what I've seen. There some other thing to click besides "add difference"?
@Trahloc聽Can some one explain how to correctly use add difference pleas ? - I use automatic1111, I know A and B but C confuses me a little, thnx in advance
@potty69聽just ran across https://old.reddit.com/r/StableDiffusion/comments/10cgxrx/wellresearched_comparison_of_training_techniques/ apparently LoRA is the way to go not hypernetworks
This would be cool as a set of textual inversion embeddings files that you could use to enhance expressions with any model.
You can extract a LORA from a checkpoint :)
@Jaxx聽Wow, I wasn't aware of that! If it works, I will upload
@terrariyum聽I think you should watch this https://www.youtube.com/watch?v=70H03cv57-o
first on how to install things, but I have seen several users here posting LORAS of other checkpoints.
@terrariyum聽thanks for the LORA version! I'll give it a try.
is there a ckpt available, I am confused on how to use the safe sensors with WebUI
No need... Put it exactly where you put your ckpt files (/models), refresh the checkpoint list and voila!
Is a pix2pix version possible?





