Krea 2 Fine-Tune Released with Coordinate-Based Comic Layout

The community has released a turbo fine-tune for the Krea 2 generative model, jimmycarter/krea2-turbo-bbox, featuring coordinate-based comic layout control, along with lvladikov's distillation LoRAs that accelerate Krea 2 Turbo from 8 to 4 and 2 steps.

Compiled by Sergey KostenchukPublished 2026-09-17Updated 2026-09-18

2026-09-17 Multimodal HuggingFace

🎨 Krea 2: Coordinate-Based Comic Layout and LoRA Distillation

Neuronavt has gathered the latest fine-tunes, LoRAs, and ComfyUI workflows for Krea 2. The centerpiece is jimmycarter/krea2-turbo-bbox: a turbo fine-tune of Krea-2-Raw for 8 steps without CFG, maintaining panels, speech bubbles, and character identity; the layout is defined by coordinates on a 0–1000 grid using the DSL from PROMPTING.md.

🌍 An open ecosystem is forming around Krea 2 in the spirit of SDXL and Flux: LoRA step distillation without retraining the base model, task-specific specialization, and open ComfyUI workflows; coordinate-based control of composition and text is a rare level of capability for open models.

👤 lvladikov's distillation LoRA accelerates Krea 2 Turbo: 4 steps — 54.5 seconds versus 88.7 seconds at 1024×1024 (~1.6×), 2 steps — ~4.2× faster denoising; measurements by the authors. For comics — krea2-turbo-bbox and PROMPTING.md.

Source 1: https://huggingface.co/jimmycarter/krea2-turbo-bbox Source 2: https://huggingface.co/lvladikov/Krea2-Turbo-Distill-4step-LoRA

Sources