Sign In

Look alive! Excited facial expression / emotion

62
1.1k
4.3k
33
Verified:
SafeTensor
Type
LoRA
Stats
1,077
4,260
156
Reviews
Published
Apr 1, 2024
Base Model
SD 1.5
Trigger Words
excited
open mouth
wide smile
wide open eyes
Hash
AutoV2
55067FC63C

What is this

Look alive! is my series of facial expression LoRAs that are intended to work together with consistent character LoRAs and embeddings. This particular LoRA is for the excited facial expression. Although there are many different facial expressions that can convey the emotion of being excited, this LoRA is trying to generate a facial expression that looks like:

  • Mouth wide open in shape similar to smiling or laughing, but without the lower teeth showing

  • Eyes either wide open (like surprised) or happy squinting (like when laughing)

  • Cheeks naturally connecting the mouth with the happy/surprised eyes

How to use it

  • During training only the "excited" trigger word was used, but when using this LoRA I recommend prompting with "excited, open mouth, wide smile, wide open eyes".

  • Start with 0.7 weight.

  • If your image doesn't have a closeup face composition, don't use this in the initial generation, but instead inpaint the face area. Use sufficient padding when inpainting (see the sample images for what composition is expected).

Why was it created

Why did I choose to train the excited expression specifically? The SD1.5 model can partially understand the "excited" expression without LoRAs - just not well enough to work with consistent characters. So it's perfectly situated between "too hard, can't make it work" and "too easy, no need for LoRA". I'll also note that I found only one pre-existing "excited" expression LoRA and it didn't give me good results when I tried it. The grid comparison contains a column for this. You can see in the grid that it's overcooked and turns photorealistic generations into very cartoonish looks (unless you actively fight against that by prompting or other LoRAs etc.). In other words, I made this LoRA because the existing alternatives were not working great for me.

For more background on my research into facial expressions on consistent characters, see my article at https://civitai.com/articles/4563/facial-expression-flexibility-with-consistent-characters