Krea 2 Daylight Look LoRA (16 GB tested)

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

🎯 Krea 2 Daylight Look LoRA v1

A warm, natural daylight look for Krea 2 portraits of adult women.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

A small style LoRA trained on 34 AI-generated portraits of adult women (Civitai on-site trainer, Krea 2 Base, 2,000 steps, epoch 6 released). It nudges portraits toward soft, directional daylight: warm sunlight, light patches on floors and walls, and crisp, detailed backgrounds, while keeping the outfit, pose and framing you prompt for. Recommended strength is 0.6.

✨ Features

• Warm directional daylight with sun patches and a more photographic feel

• Sharper, more detailed backgrounds than plain Krea 2 output

• Gentle by design: at 0.6 outfit, pose and framing stay close to your prompt

• No trigger word. Tested on a 16 GB GPU

📦 Required Models

• Base model: Krea 2 (the LoRA was trained on Krea 2 Base; the example images were made with Krea 2 Turbo)

• LoRA file: krea2_daylight_look_v1_e06.safetensors (epoch 6)

• Use the same Krea 2 files you already use for your workflow

🧩 Required Custom Nodes

• None specific. Any standard LoRA loader node that works with Krea 2 is fine

🚀 How to Use

• Step 1: load Krea 2 as usual

• Step 2: add a LoRA loader after the model loader and pick the LoRA file

• Step 3: set strength to 0.6

• Step 4: prompt in plain sentences. Naming a sunny setting helps the look show: words like sun, sunlit, golden hour or morning light work well

⚙️ Settings & Parameters

• Strength: 0.6 recommended, up to 0.8 for a stronger daylight effect

• At 1.0 pose and framing start to drift from the prompt, so it is not recommended

• Example images: Krea 2 Turbo, 8 steps, CFG 1.0, sampler er_sde_simple, 1280x1600 generation then upscaled 2x

💡 Tips

• The look shows most clearly when the prompt describes sunlight or a bright outdoor or window-lit scene. Indoor scenes with weak light change less

• If a result is not what you want, try another seed first, then lower the strength

• Not tested stacked with other style LoRAs

🔗 More from me

• My Krea 2 workflows (Turbo dual-mode, LTX motion suites) and my Qwen-Image 2.1 Glamour Realism LoRA are on my profile

📝 Notes & AI Disclosure

• All example images are AI-generated with this LoRA on Krea 2

• Training images were AI-generated too. All subjects are adult women and none are real people

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

⭐ Found this useful?

• Like if it saved you time

• Comment your results, I read every one

• Follow for new LoRAs and ComfyUI workflows, all tested on 16 GB VRAM

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

⚖️ Model Attribution & Licensing

• Built on Krea 2; this LoRA modifies the style of Krea 2 outputs

• Licensed under the Krea 2 Community License (attribution: Krea 2 by Krea)

• Not official Krea and not endorsed by Krea

• Commercial use only below USD 1M annual revenue, per the Krea 2 Community License

• This LoRA is an adapter I trained myself and contains no Krea model weights