A Wan2.1 video LoRA trained to transform live video into Dungeons & Dragons

fantasy book cover illustration style.

Trained on 40 synthetic images generated using the DnD 5e Covers Flux LoRA

(civitai.com/models/742520) as the data source, then trained on Wan2.1-T2V-1.3B

using musubi-tuner over 32 epochs.

## Usage

Use with the LongLive pipeline in Daydream Scope or any Wan2.1 compatible

inference tool.

Trigger word: dnd fantasy art style

Recommended weight: 1.2–1.6 (default 1.4)

## Example Prompts

- dnd fantasy art style, heroic warrior in ornate plate armor, dramatic lighting

- dnd fantasy art style, powerful wizard casting arcane spells, magical energy

- dnd fantasy art style, dark dungeon interior, glowing runes

- dnd fantasy art style, fearsome dragon attacking a medieval castle

## Training Details

- Base model: Wan2.1-T2V-1.3B

- Training images: 40 at 768x768

- Epochs: 32

- LoRA rank: 64

- Learning rate: 2e-4

- Trigger word: dnd fantasy art style