Instructions to use hodgy/yusmi with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use hodgy/yusmi with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("krea/Krea-2-Turbo", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("hodgy/yusmi") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
Yusmi โ Krea 2 LoRA
An identity LoRA for krea/Krea-2-Turbo, trained with Ostris AI Toolkit.
Trigger word
Use yusmi near the beginning of the prompt. The token is confirmed by the embedded tag-frequency metadata.
Example prompt:
Photorealistic portrait of yusmi, direct gaze, natural window light, detailed face, realistic skin texture, soft background blur.
Training metadata
| Parameter | Value |
|---|---|
| Base family | Krea 2 |
| Checkpoint step | 2500 |
| Recorded epoch | 16 |
| Trainer | ostris/ai-toolkit 0.12.8 |
| Confirmed trigger | yusmi |
| Tensor count | 512 |
File
yusmi_krea_2_lora_v1_000002500.safetensorsโ 2,500-step Krea 2 LoRA checkpoint
Recommended settings
- Use a compatible Krea 2 workflow/loader.
- Start around LoRA strength
0.7โ0.9; adjust toward1.0for stronger identity. - Include
yusmionce and describe framing, clothing, pose, lighting, and environment.
Limitations
Identity consistency can decrease with distant framing, occlusion, extreme angles, crowded scenes, or strong stylization. Hands, text, and small facial details may contain artifacts. The adapter may reproduce biases or recurring visual patterns from its training data.
Responsible use
Use only where you have the necessary rights and permission. Do not use this LoRA to impersonate, deceive, defame, harass, exploit, or violate privacy or publicity rights. Clearly disclose AI-generated imagery when context could cause viewers to mistake it for a real event or endorsement.
- Downloads last month
- 21