FollowCam-Redmond-WAN2-T2V-14B

Instala las dependencias y carga el LoRA sobre el modelo base Wan2.2-T2V-A14B:

pip install -U diffusers transformers accelerate

import torch
from diffusers import DiffusionPipeline
from diffusers.utils import export_to_video

# Cambia a "mps" para dispositivos Apple
pipe = DiffusionPipeline.from_pretrained(
    "Wan-AI/Wan2.2-T2V-A14B",
    dtype=torch.bfloat16,
    device_map="cuda"
)
pipe.load_lora_weights("artificialguybr/FollowCam-Redmond-WAN2-T2V-14B")

prompt = "FollowCam. A continuous wide-angle shot following a lighthouse keeper climbing the spiral staircase inside a towering coastal lighthouse during a storm. She ascends steadily, one hand on the iron railing, the other carrying a maintenance kit. The narrow stairs wind upward in dizzying circles. Wind howls outside, rain lashing against small windows. The structure groans and sways slightly. Each window reveals glimpses of angry grey ocean below. She reaches the lamp room at the top, where massive rotating lenses cast sweeping beams into the tempest."

output = pipe(prompt=prompt).frames[0]
export_to_video(output, "output.mp4")

Para usarlo en ComfyUI, descarga [WAN2.2]Followcam_Redmond_high_noise.safetensors, colócalo en ComfyUI/models/loras/ y añade un nodo Load LoRA después del cargador del modelo base. Conecta su salida MODEL al KSampler y utiliza una intensidad inicial de entre 0,8 y 1,0. Los prompts deben comenzar con «FollowCam» y describir un plano continuo que siga a un sujeto a través de un entorno.