Downloads · 30 days
0
hannahowo-krea2-lora
spyrostheboss/hannahowo-krea2-lora
hannahowo-krea2-lora is a text-to-image model from spyrostheboss. Use it when you need an image from a text prompt. The card lists the license as other.
An unofficial, community-created LoRA for Krea 2 designed to reproduce the recognizable visual appearance and facial characteristics associated with HannahOwO.
Access
Public
Updated Aug 30, 2026
Repo size
480 MB
Likes
5
Public
Repo makeup
Click a slice to open those files.
.safetensors469 MB · 100%
At a glance
- Task
- Text-to-Image
- License
- other
- Access
- Public
- Created
- Aug 23, 2026
- Updated
- Aug 30, 2026
- SHA
- ca9b1d10
Base models
Model card
From the Hugging Face model README
HannahOwO — Celebrity Likeness LoRA for Krea 2
An unofficial, community-created LoRA for Krea 2 designed to reproduce the recognizable visual appearance and facial characteristics associated with HannahOwO.
This model is not affiliated with, endorsed by, sponsored by, or officially connected to Hannah Kabel / HannahOwO.
Important: This is a third-party community model. The model author does not claim ownership of the depicted person's identity, name, likeness, or other personality/publicity rights. Users are responsible for ensuring that their own use of generated content complies with applicable laws, platform rules, and third-party rights.
Trigger Word
Use:
hannahowo
For best results, place the trigger at the beginning of the prompt.
Example:
hannahowo, portrait of a woman outdoors during golden hour, natural lighting
The trigger should be written in lowercase exactly as shown.
Recommended Settings
| Setting | Recommended Value |
|---|---|
| Base model | Krea 2 Turbo |
| Training base | Krea 2 Raw |
| LoRA strength | 0.9 |
| Steps | 8 |
| Guidance scale | 1 |
| μ (mu) | 1.15 |
| Resolution | 1024×1024 |
These settings are starting points rather than strict requirements. Results may vary depending on the prompt and Krea 2 configuration.
Training Details
| Parameter | Value |
|---|---|
| Training base | Krea 2 Raw (raw.safetensors) |
| Architecture | Krea-2 / LoRA (networks.lora_krea2) |
| Network dimension | 32 |
| Network alpha | 32 |
| Training images | 1,887 |
| Epochs | 3 |
| Training steps | 1,416 |
| Batch size | 1 |
| Gradient accumulation | 4 |
| Effective batch size | 4 |
| Optimizer | AdamW8bit |
| Learning rate | 1e-4 |
| Scheduler | Constant |
| Warmup | 0 |
| Timestep sampling | krea2_shift |
| Resolution | 1024×1024, bucketed |
| Mixed precision | BF16 |
| Base precision | FP8 |
| Seed | 42 |
| Text encoder | Qwen3-VL-4B, frozen |
| VAE | Qwen-Image VAE |
Usage
The LoRA is provided as a single .safetensors file:
hannahowo_krea2_v2.safetensors
It can be loaded using compatible Krea 2 / Musubi-Tuner workflows and ComfyUI setups that support Krea 2 LoRAs.
Disclaimer
This is an unofficial fan/community project.
It is not produced, sponsored, authorized, or endorsed by Hannah Kabel / HannahOwO or any affiliated organization.
The model is provided for research, experimentation, and creative purposes. Do not use generated content to impersonate the depicted person, falsely imply endorsement or affiliation, deceive others, or otherwise violate another person's rights.
Users are responsible for their own prompts, generated outputs, and use of the model.
License
Other / custom terms
The model is not presented as an official work of the depicted person. No rights to the depicted person's name, likeness, identity, trademarks, publicity rights, or other third-party rights are granted by this model card.
The author only grants rights, if any, to the portions of the model and documentation over which the author has the legal authority to grant them.
Please review applicable laws and the terms of the platform where the model is used before redistributing or using the model commercially.
- Task
- Text-to-Image
- License
- other
- Languages
- en
- Created
- Aug 23, 2026
- Updated
- Aug 30, 2026