Sign In

KREA2 BBOX HYBRYD W6A8

Early Access

Download

1 variant available

int8 SafeTensor

krea2-bbox-turbo-HYBRYD_W6A8.safetensors

8-bit integer, smaller file • 9.89 GB

Verified:

Type
Checkpoint Merge
Stats

13

122

Reviews
Published

Oct 28, 2026

Base Model
Hash
AutoV26C0D1282A4
00165-28773229.jpeg

KREA2 BBOX HYBRYD W6A8

Early Access — 15 Days

Experience a custom Krea2 Turbo-BBox hybrid featuring six individually trained and merged LoRAs, enhanced visual detail, cinematic lighting, photorealism, fantasy, sci-fi and artistic versatility.

Features

  • Custom hybrid based on Krea2 Turbo-BBox EMA 0.99975

  • 6 personally trained LoRAs

  • W6A8 quantization

  • Optimized for 6–8 sampling steps

  • CFG 1.0

  • Tested with Forge Neo

  • BBox spatial prompting support

  • Your Buzz support helps fund future model experiments, training and development.

Thank you for supporting independent AI model creation!

Credits & Original Model

KREA2 BBOX HYBRYD W6A8 is an independent custom hybrid checkpoint based on Krea2 Turbo-BBox EMA 0.99975, created by jimmycarter.

Original BBox model:
https://huggingface.co/jimmycarter/krea2-turbo-bbox

Original Krea2 architecture:
https://huggingface.co/krea

Official Krea 2 Community License:
https://www.krea.ai/krea-2-licensing

This derivative includes six custom LoRAs trained by me using my own image datasets, followed by model merging and W6A8 quantization.

Special thanks to jimmycarter for the original BBox fine-tune and to the Krea team for the base model.

This is an independent community project, not an official Krea release.