Stable Diffusion · LoRA

FishEye Redmond WAN2 T2V 14B

LoRA for Wan2.2 T2V 14B that adds a wide-angle fish-eye lens distortion effect to text-to-video generations.

About this model

FishEye Redmond WAN2 T2V 14B is a LoRA (Low-Rank Adaptation) model trained on the Wan Video 2.2 T2V-A14B base model. It specializes in applying a fish-eye lens effect to text-to-video generations, producing a characteristic convex, non-linear wide-angle distortion where the center of the frame appears closer and the edges are more stretched. This creates a distinctive, curved visual style suited for artistic, experimental, or dramatic visual storytelling.

The model is distributed as two safetensors files: a low-noise variant and a high-noise variant, both intended to be used together in ComfyUI or compatible diffusers workflows. The recommended setup places both files in the ComfyUI/models/loras/ directory and applies them after the base Wan model loader, with strengths between 0.8 and 1.0 for the model parameter.

The trigger phrase is "Fish eye effect" — append it to the end of your prompt to activate the distortion. The model ships with several example prompts covering dance, book scenes, soap bubbles, cycling, and portraiture. Training was made possible by GPU time sponsored by Redmond Ai, and the model is released under the CC0 1.0 license, allowing free commercial use without restriction.