The Look – A LoRA for Natural Faces & Expressions
Trigger keyword: LookDaal
Model type: LoRA (Flux)
🌟 What is The Look?
The Look is a LoRA designed to break away from the lifeless “AI dead stare” and unlock a wide spectrum of expressive, natural, and meaningful faces. It’s built to give your portraits more personality, with varied gazes, expressions, and subtle emotional depth.
To do this (it took me well longer than it should have to figure out how to share the wildcards.) I have a set of randomization prompts to help create great images. The Look Wildcards - v1.0 | Flux Wildcards | Civitai
This LoRA is ideal for:
Creating characters with authentic, lived-in looks
Exploring diverse glances and eye contact styles
Adding subtle nuance and emotional storytelling to portraits
Expanding your options for face shapes, moods, and angles
⚙️ How to Use
Activation keyword:
LookDaalRecommended strength:
1.0(works consistently well at this level)Built for Flux (future versions will support other base models).
Pairs nicely with other LoRAs for style, lighting, or character design. (So far I’ve tested mainly with my own, but it should combine well with others.)
🎨 Features
Rich set of facial expressions (from soft smiles to focused intensity)
Wide range of eye positions & gazes (direct, sideways, wistful, playful, etc.)
Natural face and head tilts, postures, and angles
Works seamlessly with different styles (realistic, painterly, anime-inspired, etc.)
Helps avoid the dreaded blank, lifeless look in portraits
📌 Notes
Best results when used with prompts that highlight expression, gaze, and emotion.
Consider layering with other LoRAs for wardrobe, backgrounds, or lighting to push character depth even further.
If you experiment with it in combination with other LoRAs, I’d love to hear about the results!
✨ This is the first release of The Look, tuned for Flux. More versions for other models are planned once this foundation is set.
Description
Trained with Qwen. This is my first try with Qwen but it does seem to have a marked impact vs. vanilla. It combines well with Qwen-image-lighting and has proven very flexible in output format while holding to prompts. Workflow is embedded in images with working Qwen, randoms, and The Look Wildcards - v1.0 | Wildcards
FAQ
Comments (8)
Even if it is just your second lora ported to ZIT my conclusion is already >impressive<
Thank you for the kind words. I am still learning a little bit each time but I do have to say this is quickly becoming my favorite model to train against. It does a few wackado breaks now and then from intended operation but for the most part it behaves really well. I look forward to soon launching new ones that I start with ZIT.
This lora is poorly named. The effect it has on images made by zit far exceeds only facial expressions. It has a positive effect on the entire image. This is great.
Thank you. The Look was originally targeted to get out of the "dead stare". It turns out that sourcing those well made images also infused a set of other positives. Thanks again for the kind words and comment.
I second this comment! This LORA is great! It is the first LORA on ZIT that I know that ADDS to the image. It doesn't interfere altering my styles, the character and what I described in the prompt. But it adds something. A quid. Still testing but when I compare my pleasing image without it to the one with it I prefer the second one. Not the facial expression. Atmosphere, details. Something in the air... BTW FYI: doing anime/cartoonish and I use it at very low strength: .25 seems my sweet spot. Kinda a spice I need in a plate to make it complete.
@BiTZero Awesome! Thanks so much for the feedback! I'm so glad you like it.