Krea 2 Lip Biting LoRA (v2)
A concept LoRA for Krea 2 that adds a natural lip-biting expression: the upper teeth resting on or gently pressing into the lower lip. It changes the mouth and leaves the rest of the image alone. Faces, hair, clothing, lighting and composition stay close to what the base model would have made without it, so it combines well with character and style LoRAs.
Usage
- Trigger word:
lpb1te. Put it anywhere in the prompt; the examples put it at the end. - Strength: 0.8β1.0. Around 0.6 the effect is a light tuck of the lower lip; 0.8β1.0 gives a clear bite; at 1.2 it is firmer and the expression can start to shift.
- Base models: trained on Krea 2 Raw. It also works on Krea 2 Turbo-based checkpoints and merges.
- With character LoRAs: a character LoRA at 0.8 alongside this one at 0.8β1.0 keeps the character's face and still shows the bite.
- Prompting: describe a relaxed or quiet expression and let the LoRA supply the bite. Prompts that ask for a big smile or an open mouth will fight it.
Example prompt
Upper-body portrait photograph of a woman in her mid-twenties with long strawberry-blonde hair in loose, tousled waves, a few strands falling across her cheek, the rest tucked behind one ear. Grey-green eyes, a cute youthful face with delicate features and a light scattering of freckles across her nose and cheeks. Fair, smooth skin; minimal makeup that highlights her natural features. She wears a deep wine-red lace plunge-neck top that frames her collarbones, with a fine gold chain necklace. Relaxed, quiet expression, looking at the camera. She stands against a plain, light-toned studio backdrop. Modern fashion-editorial style. 2/3 dynamic camera angle. Shot on an 85mm lens wide open at f/1.4, razor-thin depth of field, creamy circular bokeh. Warm low-key studio light from the side, golden rays catching her hair, gentle falloff into shadow, golden-amber highlights, deep warm blacks, dim volumetric contrast. Kodak Portra 400 film look, fine organic grain, slight halation around highlights, muted color palette, soft rolled-off highlights. Tactile textures, fabric and skin rendered in fine sharp detail. Dreamy, sensual mood, quiet and private atmosphere. lpb1te
Seed 20261108, 1184Γ1760.
Training
The LoRA was trained with ai-toolkit on Krea 2 Raw (bf16). The approach is built around keeping the effect on the mouth:
Data: captioned portraits showing a lip bite, each with a hand-reviewed mouth mask, in three groups:
- a larger set of varied faces, repeated 3Γ
- an older set, repeated 1Γ
- a small set of especially clear, strong bites, repeated 1Γ
The groups are pooled and shuffled together. The mask files sit inside each group's folder, so the trainer also sees them as images.
Masked loss: the mouth mask concentrates learning on the lips and teeth. The rest of the image still contributes at a 0.2 floor, so the model keeps seeing the whole face.
Inverted-mask prior (0.5): outside the mouth, the LoRA is pulled toward the base model's own prediction. This is what keeps it from learning the training set's faces, so identities in your prompts and character LoRAs are preserved.
Captions: used as written, with the trigger word kept in place and 5% caption dropout.
| Setting | Value |
|---|---|
| Network | LoRA, rank 16, alpha 16 |
| Steps | 2,000 |
| Batch size | 1 |
| Optimizer | AdamW 8-bit, learning rate 1e-4 (constant) |
| Noise schedule | flow matching, linear timesteps |
| Resolution | 1024, aspect-ratio buckets |
| Precision | bf16 (base model quantised for training) |
| Hardware | 1Γ NVIDIA H200 |
The record of the run is in training/:
training-settings.json: the training settings, the workflow package and trainer image they ran on, the dataset at an exact commit, and the base model file's SHA-256.- Pins the workflow package
krea2-lorav4, commitcc95424e, and the trainer imageostris/aitoolkit@sha256:d0b5f729β¦(ai-toolkit 0.13.41).
- Pins the workflow package
ai-toolkit-config.json: the ai-toolkit config that produces these weights.
Dataset: abide1622/lip-biting-V2 at commit 811aa3bf, subsets main, legacy and strong (94 images).
Images
Strength 1.0, no character LoRA.
Rows: no character LoRA (None), then three different character LoRAs (A, B, C) at strength 0.8, to show how the bite combines with an identity. Columns: LoRA strength 0, 0.6, 0.8, 1.0 and 1.2. Same prompt and seed (20261108) in every column of a row.
Rendered on a Krea 2 Turbo-based photoreal merge at its usual settings: 8 steps, euler sampler, simple scheduler, CFG 1, 1184Γ1760.
Limitations
- Use it in prompts that show a face. With no face in the scene, the trigger word can end up drawn as text (on a sign, for example).
- The training images are portraits of adult women. The effect works best on similar framing (head-and-shoulders to upper body) and may be weaker on wide shots or other subjects.
- It shapes an expression. It is not meant to change who the person is; if a face drifts, lower the strength.
License
- Krea 2 Community License. This LoRA is a modification of the Krea 2 model, so it is distributed under that license. A copy is included as
LICENSE-Krea-2.pdf, and the required notice is inNOTICE. By using this LoRA you agree to its terms. This is not an official Krea product and is not endorsed by Krea. - Non-commercial. The training dataset is licensed CC BY-NC 4.0, so this LoRA is offered for non-commercial use only.
Model tree for abide1622/krea2-lora-lipbiting-v2
Base model
krea/Krea-2-Raw
