Hugging Face's logo

UntMods
/
Krea2_Chars_LoRA

Text-to-Image
lora
krea2
character
comfyui
Model card Files
xet
 Community
1
Krea2_Chars_LoRA / README.md
PGC
Add model card README with updated contents table
fec1489
preview code
|
Raw
History Blame Contribute Delete
4.26 kB
metadata 
license: apache-2.0
tags:
  - lora
  - krea2
  - character
  - comfyui
  - text-to-image
base_model: KREA2

Krea2 Character LoRA Collection

A collection of 34 character LoRAs for the KREA2 base model, each paired with a persona .md file describing the character's stable facial features, variable appearance, and expressions.

Every LoRA ships as <Name>_KREA2.safetensors alongside <Name>_KREA2.md.

Fully automated pipeline

This entire repository — dataset, training, and upload — was produced end-to-end by an autonomous Hermes Agent (served by the Kimi K3 model) from a natural-language request. No manual review gates anywhere in the loop:

  1. Scraping — automated image collection per character from web photo pools.
  2. Cropping & curation — face/head detection Ultralytics Yolo and crop, with a hard content filter (No AI generated images, correct subject only, sharpness threshold).
  3. Aesthetic gating — every candidate image scored by a discriminator built on the ERNIE aesthetic image scorer; low scorers rejected automatically with a hard pass/fail threshold.
  4. Captioning — datasets captioned with Gemma-4 26B (abliterated) vision MTP for unrestricted, descriptive captions, fast (1.2s/i).
  5. Persona synthesis — each <Name>_KREA2.md generated from caption trait-analysis (facial features, range of expressions).
  6. Infrastructure — training pods created, configured, monitored, and torn down automatically via the RunPod API (MCP).
  7. Training — LoRAs trained on pods with Ostris AI-Toolkit with user predefined settings.
  8. Publishing — weights and persona files uploaded to this repo by the agent.

Usage

Recommended: CRT-Nodes persona loader

Use the Seeded Persona LoRA Loader (CRT) node from CRT-Nodes. It loads both the LoRA and its paired .md file in one node, and returns:

  • MODEL — the LoRA-patched model
  • STRING — the persona text, ready to feed an LLM prompt enhancer

Wiring the string output into your prompt-enhancement LLM gives it grounded hints about the currently loaded character — stable features, typical styling, expressions — so prompts stay on-identity without you memorizing trigger details.

The same CRT-Nodes pack includes ERNIE Image Aesthetic Score (CRT) — the exact scorer used to gate these datasets, packaged as a ComfyUI node with automatic model download. Ready to use out of the box if you want to score your own outputs or curate your own datasets with the same standard.

Standard usage

Load <Name>_KREA2.safetensors as a regular LoRA on top of KREA2 in ComfyUI (or any KREA2-compatible pipeline) — recommended strength is around 1.0. The .md files are plain text — read them, or feed them to your prompt LLM manually.

Contents

Adriana Lima Alexandria Ocasio-Cortez Ana De Armas
Anne Hathaway Anya Taylor-Joy Ariana Grande
Autumn Falls Bella Poarch Billie Eilish
Camille Cerf Doja Cat Dorozea
Dua Lipa Elydia Emilia Clarke
Eva Mendes Hailey Rose Hannah Kleit
Ho-yeon Ice Spice Inde Navarrette
Jessica Alba Kaia Gerber Kylie Jenner
Lea Elui Nicki Minaj Sab Zada
Shakira Shay Sydney Sweeney
Tinashe Tokyo Zendaya
Zoe Kravitz

Known limitations

  • Trained for likeness; style flexibility depends on your prompt and sampler settings.

Disclaimer

These LoRAs reproduce the likeness of real people and are shared for research and creative purposes. Please use them responsibly and respect the rights of the individuals depicted — don't create misleading, harmful, or non-consensual content. By downloading, you take responsibility for what you generate.

Credits