README.md

---
title: Krea 2 Style Reference
emoji: 🎨
colorFrom: indigo
colorTo: blue
sdk: gradio
sdk_version: 6.20.0
app_file: app.py
pinned: false
hardware: zero-a10g
short_description: Style-guided image generation with Krea 2 Turbo
python_version: "3.10"
startup_duration_timeout: "30m"
models:
  - krea/Krea-2-Turbo
  - ostris/krea2_turbo_style_reference
---

# Krea 2 Style Reference

Generate an image from a text prompt, guided by a **style reference image**. This Space uses:

- **Base model:** [krea/Krea-2-Turbo](https://huggingface.co/krea/Krea-2-Turbo) (8-step distilled Krea 2)
- **LoRA:** [ostris/krea2_turbo_style_reference](https://huggingface.co/ostris/krea2_turbo_style_reference) (style reference conditioning)
- **Pipeline:** [ostris/Krea2OstrisEdit](https://huggingface.co/ostris/Krea2OstrisEdit) (custom community pipeline for reference-image conditioning)

## How it works

Upload a style reference image and type a prompt describing what you want to generate. The LoRA conditions the Krea 2 Turbo model on both:
1. **Text encoder** — the reference image is embedded into the prompt via Qwen3-VL vision placeholders
2. **Latent space** — clean VAE latents of the reference are appended to the transformer sequence at flow time t=0

This produces an image that follows your prompt while adopting the visual style of the reference image.

## Inference recipe

- **Steps:** 8 (Turbo distilled)
- **Guidance:** 0.0 (disabled, per Krea 2 Turbo convention)
- **LoRA scale:** 1.0 (adjustable)
- **Resolution:** 1024×1024 (adjustable, must be multiples of 16)