Stable Diffusion · LoRA

WindowSeat Redmond — Wan 2.2 T2V 14B LoRA

Wan 2.2 LoRA that renders surreal, cinematic views from bus, train and plane windows during travel.

About this model

WindowSeat is a LoRA for Wan 2.2 T2V A14B that captures the contemplative, surreal experience of looking through windows while traveling. Each generated scene is framed as a view from a bus, train, plane, subway or tram that reveals something extraordinary: floating islands above clouds, an orchestra performing in a wheat field, a spotlit giraffe on a theater stage, or a desert turning into mirror-like salt flats. The style combines cinematic framing with surreal realism, turning the mundane act of looking out a window into a portal to other worlds.

To use it, start your prompt with the trigger phrase "Window Seat" followed by a description of a view through a vehicle window that reveals something surreal. The model card recommends describing the camera movement (e.g., handheld wide shot from inside a moving bus, locked-off shot from a subway) alongside the scene, and suggests film grain and color-grading cues such as "warm 16mm film grain" or "Tarkovsky's contemplative surrealism" for a specific look.

In ComfyUI, download [WAN2.2]WindowSeat_Redmond_high_noise.safetensors into ComfyUI/models/loras/, add a Load LoRA node after your Wan base model loader, and connect the MODEL output to the KSampler. The model card recommends a strength_model of 0.8–1.0 as a starting point. The base model is Wan-AI/Wan2.2-T2V-A14B, the LoRA is released under the CC0 1.0 license, and training was sponsored by Redmond AI.