Krea 2 AI Image Model in Stable Diffusion Forge Neo — A ComfyUI Alternative
Guide

Krea 2 AI Image Model in Stable Diffusion Forge Neo — A ComfyUI Alternative

Jul 2026 · 6 min read · Krea 2 · Forge Neo · GGUF · FP8 · Text to Image

Krea 2 is an image model built for fast, high-quality generation with a strong focus on realism, prompt accuracy and style control — it produces clean, detailed images with very little prompt engineering. Stable Diffusion Forge Neo now supports it fully, so you can run Krea 2 from a simple web UI instead of ComfyUI.

Why Forge Neo for Krea 2

Forge Neo is the continuation of the last Forge release. It runs SD, Flux, WAN 2.2, Z-Image Turbo and Krea 2 in one lightweight interface, with performance and simplicity as the focus — no node graph to wire up. If you haven't installed it yet, follow how to install Forge Neo on Windows.

Model Files

FileWhat it's for
krea2_turbo-Q3_K_M.gguf (or another GGUF quant)Krea 2 Turbo — low VRAM
krea2_turbo_fp8.safetensorsKrea 2 Turbo FP8 — 12 GB+ VRAM
qwen3vl_4b_fp8_scaled.safetensorsText encoder
qwen_image_vae.safetensorsVAE
VRAM guide: at least 6 GB for the GGUF models and 12 GB+ for FP8, on an NVIDIA RTX 30-, 40- or 50-series card. Leave at least 20 GB free for the models.

Using It

  1. Make sure Forge Neo is up to date — Krea 2 support needs a recent version.
  2. Place the model, text encoder and VAE in Neo's model folders and select them at the top of the UI.
  3. Add any Krea 2 LoRAs under the LoRA tab.
  4. For text-to-image, confirm the correct models are selected, then generate.

FFmpeg should be installed too.

Prefer ComfyUI? Free Krea 2 Workflows

Since this video, several free Krea 2 workflows have been released for ComfyUI:

📦 Want to skip the setup?

The Krea 2 Forge Neo installer sets up Sage Attention 2, Flash Attention 2 and Triton, and downloads the GGUF model, text encoder and VAE for you.

Get the Installer →