Stable Diffusion · LoRA

Bloom Redmond LoRA for Wan2.1 T2V 14B

LoRA for Wan2.1-T2V-14B that renders surreal, cinematic floral sculptures blending technology, art, and nature via the trigger phrase 'Bloom'.

About this model

This LoRA adapts the "Bloom" artistic style for the Wan-AI/Wan2.1-T2V-14B text-to-video model. The aesthetic centers on large-scale, impossible floral sculptures that merge mechanical and natural elements into cinematic, macro-timelapse compositions. Trained with GPU sponsorship from Redmond Ai, the LoRA was built to extend the base video model with a surreal sculptural vocabulary rather than alter its underlying mechanics.

The trigger phrase is "Bloom". To use it, prepend the word to a prompt describing a grand, surreal floral sculpture or installation, and the model renders it in the signature aesthetic. Example generations from the model card span brass-gear clockwork buds, children's clay and crayon wax petals, LEGO brick formations with transparent and Technic elements, and wet porcelain clay shaped on a potter's wheel with cobalt-blue slip cores. Each example prompt describes a macro timelapse or slow cinematic reveal where petals unfold to expose intricate internal mechanisms, glowing vacuum tubes, or refracted light.

In ComfyUI, install the LoRA by downloading Wan-Bloom-Redmond.safetensors (292.6 MB) and placing it in ComfyUI/models/loras/. After loading the Wan2.1-T2V-14B base model, add a Load LoRA node connected to the model output, select the LoRA file, and route its MODEL output to the KSampler. A strength_model value of 0.8–1.0 is recommended as a starting point. Ensure the ComfyUI environment is already configured to run Wan video models.

The model is released under the CC0-1.0 license, placing it in the public domain with no usage restrictions. It is compatible with both text-to-video and text-to-image generation through the diffusers library.