Day & Night Slider | Krea2 [Zero-Train]

🤝 Support & Custom Models

💠 Sponsorship: For sponsorship, please contact us on . Your support helps us grow and improve future work.

💠 Buy Custom Models: If you're interested in buying a LoRA model, —we can get it done in a short time.

Day & Night Slider | Krea2 [Zero-Train]

A high-potency closed-form lighting slider for Krea2. Seamlessly shifts scene illumination between midday daylight and pitch-black night without corrupting character likeness, wardrobe patterns, or background geometry.

Recommended Weights

• +0.3 ~ +0.6: Dusk, twilight, and evening mood lighting. Softens ambient skylight and introduces subtle shadows.

• +0.7 ~ +1.0 (Full Night Sweet Spot): Complete nighttime conversion. Backgrounds drop into darkness, ambient daylight vanishes, and clean subject lighting/night photography dynamics engage.

• +1.2+: High-contrast dramatic night / heavy flash portrait aesthetic (best tested per prompt).

• -0.5 ~ -1.0: Inverses night or moody scenes back into clear, bright daytime illumination.

Note on Model Sensitivity: Because zero-train vectors are mathematically exact, this slider is very sensitive. On base Krea2, +0.7 to +1.0 is usually all you need for a full night conversion. If you are using community fine-tunes or heavily merged checkpoints, start at +0.5 and adjust up or down depending on the checkpoint's baseline lighting bias.

Prompting Tip

You do not need to add "night", "dark", or "moon" to your positive prompt. The slider handles the environmental lighting shift entirely through the model weights.

What is Zero-Train?

This slider was solved using closed-form analytical weight solving, bypassing traditional LoRA dataset training completely:

• Extracts the semantic direction vector between day and night concepts directly via the text encoder.

• Solves the low-rank delta matrices instantly using linear algebra.

• Isolates global lighting and environmental exposure while leaving facial features, clothing details, and composition intact.

• No iterative training steps. No dataset baking. Pure closed-form math.

👉 ZeroTrain-Slider on GitHub

⚠️Disclaimer:

This is a fictional AI-generated model created for artistic, creative, and entertainment purposes only. It does not represent any real person, living or deceased, and contains no realistic depictions of actual individuals or celebrities. The model is not intended to generate content involving real people, celebrities, or realistic human likenesses. By using this model, you agree not to generate any NSFW, harmful, or illegal content, especially involving real individuals or celebrities. Use responsibly and in accordance with ethical AI practices.

💬 Feedback & Support

If you appreciate the work I’ve put into this model, I kindly ask you to take a moment to leave feedback. Your input is invaluable and will help me improve future project