Model annotation: Please comply with relevant laws and regulations. This is for learning and research purposes only. The generation and dissemination of illegal content are prohibited.

Bilibili/YouTube/Runninghub：在下鱼酥

RunningHub Invitation Code: rh-v1525, new users receive 1000 coins, plus 100 coins daily.
|CN Invitation Code:
https://www.runninghub.cn/?inviteCode=rh-v1525
AI Invitation Code:
https://www.runninghub.ai/?inviteCode=rh-v1525
鱼酥AI website:
https://vibex.runninghub.cn/p/app-b3d967d5d3d44cc8a9ddd789af586cbc/?inviteCode=rh-v1525

Triple-track ultimate raw image workflow (cn):
https://www.runninghub.cn/post/2079122786253103105/?inviteCode=rh-v1525
Triple-track ultimate raw image application (cn):
http://runninghub.cn/ai-Tail/2079132768721793025/?inviteCode=rh-v1525
Triple-Anatomical Image Generation Workflow (AI):
https://www.runninghub.ai/zh-cn/post/2078092662537146369/?inviteCode=rh-v1525
Triple-Anatomical Image Generation Application (AI):
https://www.runninghub.cn/post/2079122786253103105/?inviteCode=rh-v1525

## Krea 2 | Cute and Sweet Portrait LoRA

This LoRA model is trained on Krea 2, primarily enhancing the cute, sweet, and trendy features and soft, approachable feel of the character. Generated characters typically possess full, smooth facial contours, soft jawlines, natural apple cheeks, clear and bright eyes, and delicately proportioned features.

The model emphasizes a natural, clean, and youthful feel, rather than an exaggerated cartoonish effect. It retains the realistic feel of the character while creating a more rounded, soft, cute, and approachable visual effect.

Suitable for:

* Realistic and semi-realistic portraits

* Fresh and natural lifestyle photos

* Fashion and beauty portraits

* Social media avatars and character design

* Japanese, Korean, and soft-focus photography styles

* Close-ups, half-body shots, and facial close-ups

### Recommended Settings

**Recommended Resolution: 720*1280**

The training dataset's image composition is not 720*1280, so it needs to be used in conjunction with my workflow.

It is recommended to start testing with LoRA weights from 0.6–0.9. Lower weights result in a more natural, slightly cute and sweet effect; higher weights further enhance rounded contours, apple cheeks, and a sweet and approachable feel. Actual results will be affected by cue words, base model, sampling parameters, and composition.

### Recommended Keywords

Adult portrait, youthful facial proportions, soft round face, smooth jawline, full cheeks, natural apple cheeks, delicate facial features, bright eyes, soft expression, clean skin texture, balanced facial proportions, close-up portrait, soft lighting, high detail

### Usage Suggestions

It is recommended to explicitly include descriptions of adult figures such as adult/adult woman/adult man in the keywords to achieve a clear age positioning and a natural, healthy portrait result. Combining this with soft lighting, natural makeup, a clean background, and medium to close-up composition will generally better showcase the model's core features.