KXSR Traveling LorA for WAN 2.1 14B

In collaboration with @machinedelusions [https://civarchive.com/user/machinedelusions]

KXSR Labs presents:
This LoRA enables the creation of cinematic sequences featuring characters and imaginative interiors with motion in the background or outside the room (typically a window if using our prompt styling). The typical view will be from inside a moving vehicle or structure with the exterior world being dynamic movement in horizontal aspect ratio generation. Can still create fun results at other aspect ratios/resolutions.

Use the trigger word "kxsr" to activate the model's specialized training.

Prompt Format

kxsr, [describe the person/thing and interior], outside the [describe the movement or happenings outside in the world around them]

Example Prompt

kxsr, A cartoon raccoon reclines in a cozy camper, wearing a yellow hoodie and jeans, with hands behind its head. A laptop, headphones, book, and mug sit on the table. Outside, trees with autumn leaves whiz by, creating a sense of motion and adventure. Twinkling lights and hanging plants decorate the camper.

Recommended Settings

• CFG: 5.5

• Shift value: 4.0

• LoRA strength: 1.0

• ~73 frames

• 720x1280 horizontal aspect ratio

Technical Details

• Base model: WAN 2.1 14B Text2Video

• Training dataset: 80 clips

• Resolution: 1280 x 720 (horizontal format)

• Frame count: 73 frames per clip

• For optimal results, maintain these specifications during inference

This LoRA works best when you provide detailed descriptions of both the subject and the surrounding environment while following the prescribed format.

Workflow Available here:
https://drive.google.com/file/d/1qrCm_iA9UNwqjzG8YRRa66a6O39F5afw/view?usp=sharing
Screenshot shows my typical inference testing setup for lora evals: