Krea 2 Moodboard + Identity Edit + Re-render for Forge Neo

Three features for the Krea 2 (K2) model in SD WebUI Forge Neo, built on a full native Qwen3-VL vision-encoder integration.

On Civitai (release zips + showcase): civitai.com/models/2794961 · ComfyUI sibling: ComfyUI-Krea2Moodboard

⚡ Plug and play: one backend patch + three extensions, everything in the normal txt2img UI — no ComfyUI, no node graphs, no dependency stack.

v3.1: adds Krea2 Re-render (Anything2Real and other ai-toolkit edit LoRAs), same patch.

v3.0, October 2026: Forge Neo 2.28 added its own Krea 2 edit and vision support, which broke the July patch. The patches are rebuilt on top of Neo's native support: the full patch needs Neo 2.29.1 or newer, an Identity Edit only patch covers 2.28 to 2.29.0, and the July bundle lives in legacy/ for older builds. See INSTALL.md.

🎨 Krea2 Moodboard

Drop reference images into a gallery — generations inherit their style / vibe (like krea.ai's Moodboard). Training-free. Controls:

• Vibe strength — 1.0 raw reference, lower = purer extract
• Vibe extract — style (palette/texture/mood; subjects fade) or subject (composition survives, style is whitened away so your prompt controls the look)
• Reference processing — full image / quadrant crops / fine 4×4 tiles (tiles = strongest composition-scrambling, style-only transfer)
• Style directive, Indirect vibe (refs hidden from the image model — structurally cannot copy pose/subject), image tokens before/after prompt
• Multiple references are packed into ONE vision span → they blend into a joint vibe (and can't trigger grid/collage outputs)

👤 Krea2 Identity Edit

Instruction-based, identity-preserving editing via community krea2_edit LoRAs (krea2_identity_edit, weights also on HF conradlocke/krea2-identity-edit). Port of ComfyUI-Krea2Edit: dual conditioning — VAE source tokens at RoPE frame 1 (appearance) + image-grounded Qwen3-VL instruction encoding (semantics), grounded negative for CFG > 1, grounding_px likeness↔obedience dial, two-ref support, aspect-ratio handling (match source / crop source to your AR / fit).

v1.2 dials (matching the LoRA's v1.2 release):

• ref_boost — reference-fidelity dial: multiplies target→reference attention (1.0 = off; the LoRA author suggests 2–6). Separate ref_boost (scene) slider for the first image in two-ref setups.
• fit source to output (v1.2) AR mode — the source is fitted in pixel space to your output resolution before VAE-encoding: blur-proof (latents are never resized), keeps your chosen AR (no more matching the source's), and matches the v1.2 LoRA's training geometry. The older match/crop modes remain for v1/v1.1 weights.
• Auto face-ref prep (2-pass) — one toggle that rebuilds the subject reference before editing: pass 1 extracts the clean, unobstructed person (hats/glasses/hands/props removed, matte natural skin), pass 2 makes a front-facing identity headshot from it. Cached per reference image (content hash) so it never re-runs for the same picture.

The two compose: enable both to take identity from the edit source and style from the moodboard.

🔁 Krea2 Re-render

The picture is rendered again from itself through an ai-toolkit Krea 2 edit LoRA, for example Anything2Real: an illustration goes in, a photograph comes out. It follows the Anything2Real ComfyUI workflow step by step and matches its results:

• the source is desaturated, cropped and sized to the output (longest side 1536, sides rounded to 512)
• Qwen3-VL reads it at the Vision size (384 by default) under the workflow's own system instruction
• the reference is conditioned at t=0 the way ai-toolkit trains these LoRAs: KV cache mode for LoRAs trained with kv_cache (Anything2Real), t=0 in sequence for the others
• recipe sampling in one checkbox: Euler, Beta 0.5 / 0.7 (beta57), 8 steps, CFG 1
• in img2img it re-renders the img2img picture, and denoise below 1 keeps part of it

Stack your other LoRAs in the prompt next to the conversion LoRA as usual.

Requirements

1. Forge Neo (neo branch), 2.29.1 or newer for the full bundle. INSTALL.md lists the patch for older builds.
2. Qwen3-VL-4B text encoder with vision weights in your Text Encoder dropdown: Comfy-Org/Krea-2 → text_encoders/qwen3vl_4b_bf16.safetensors (the fp8_scaled variant also works).
3. For Identity Edit: a krea2_edit LoRA at strength 1.0 (download from its civitai page, not bundled).
4. For Re-render: an ai-toolkit Krea 2 edit LoRA such as Anything2Real (not bundled).

What's in this bundle

Item	What	Install
extensions/	the Moodboard, Identity Edit and Re-render extensions	copy into <forge>/extensions/
krea2-features-backend.patch	4-file git apply patch for Neo 2.29.1+: the Moodboard (packed references, native Qwen3-VL processing, strength / extract / indirect) and Identity Edit (ref_boost, fit geometry, grounding_px, grounded negative) and Re-render (t=0 references in KV cache or in-sequence mode, system instruction, vision sizing) on top of Neo's native Krea 2 support	git apply from <forge>
krea2-identity-edit-backend.patch	2-file Identity Edit only patch for Neo 2.28 to 2.29.0	git apply from <forge>
legacy/	the July 2026 bundle for Neo before 2.28	see INSTALL.md

See INSTALL.md for step-by-step instructions.

Quick settings reference

• Moodboard "Krea vibe": extract style, strength 0.5, fine tiles 4×4, directive on, position after
• Identity Edit: Euler / Simple; Turbo 8 steps CFG 1 (most edits; v1.2 LoRA: 8–12 steps — 8 favors composition, 12 face detail) or Raw 20–40 steps CFG 3 (removals); ≤2MP; grounding_px 768 (1024+ for people); with the v1.2 LoRA use AR mode fit and try ref_boost 2–6
• Re-render: the defaults are the recipe; LoRA at 1.0, keep "Use the recipe sampling" on
• Prompting edits: describe only what changes; anchor with "this person"; one edit per pass

Credits & License

• Haoming02 — SD WebUI Forge Neo (this bundle's host application)
• ComfyUI — Qwen3-VL model code ported from qwen35.py / qwen3vl.py / llama.py (source URLs in file headers, following Forge Neo's existing porting convention)
• lbouaraba/ComfyUI-Krea2Edit (Apache-2.0) — the identity-edit dual-conditioning recipe and DiT in-context forward this port implements; and the krea2_identity_edit LoRA
• ethanfel (ComfyUI-Krea2TextEncoder) & ostris — validated K2 vision-conditioning recipes
• ostris/ComfyUI-Krea2-Ostris-Edit: the t=0 and kv_cache reference conditioning Re-render follows; ComfyUI-Apt_Preset's Easy_QwenEdit2509: the system instruction and vision sizing the Anything2Real workflow uses
• Krea.ai — Krea 2 (weights under the Krea 2 Community License; the LoRA is a Derivative Model — see its page)

Code in this bundle is distributed under the same license as SD WebUI Forge Neo (AGPL-3.0). Not affiliated with Krea.ai, Haoming02, or the LoRA author.