TOT_Halloween

TOT Halloween Collection - Multi-Architecture LoRA

This page presents a versatile collection of LoRA models designed to bring a distinct Halloween atmosphere, gothic details, and thematic aesthetics to your generations. The models were trained on a carefully curated and captioned dataset, delivering a strong balance between dark horror scenes and warmer autumn compositions.

The collection covers several core style directions. First is gothic digital art, featuring detailed depictions of graveyards, ancient castles, reapers, horse-drawn carriages, and eerie silhouettes under moonlight. Additionally, the model excels at autumn decorative setups such as carved jack-o'-lanterns, scarecrows in wheat fields, and warm orange lighting. It is also capable of generating realistic photography of physical outdoor decorations and masks, alongside conceptual portraits with intricate Day of the Dead style makeup.

Usage Recommendations and Parameters

• Trigger Word: tot_halloween

• Recommended Weight for SDXL, Pony, and Recommended versions: 0.7 to 0.9

• Recommended Weight for ZETA Turbo: 0.8 to 1.0

• Recommended Weight for Flux 9B Klein Dev: 1.1 to 1.3 (due to the higher rigidity and base prompt adherence of the Flux model)

• Useful Prompt Keywords: spooky atmosphere, glowing jack-o'-lantern, dark misty night, eerie lighting, dramatic gothic makeup

Available Versions in the Collection

To ensure maximum compatibility across different pipelines, the collection is split into several targeted variants:

1. Recommended Version: The core and most versatile option, optimized for a wide variety of SDXL base models.

2. ZETA Turbo: Designed for fast generations with higher contrast and sharp details.

3. Krea 2: Focused on cinematic depth and warm, vibrant orange tones.

4. Flux 9B Klein Dev: A dedicated build configured specifically for users working within the Flux framework.

Community Support

Preparing dataset captions, fine-tuning, and running test generations across multiple model architectures requires significant time and platform resources.

If you enjoy using these models and find them helpful in your workflow, any Buzz donations or image shares in the gallery below directly help cover generation costs and support the release of future resources for the community.