关于Wan2.2 TI2V模型只能生成横屏视频

设置是这样的：
（1）配置1
tarh, tarw = 224, 384
imglist = VideoData( image_folder=rootr, height=tarh, width=tarw, file_list = idlist[mi] )
（2）配置2
tarh, tarw = 384, 223
imglist = VideoData( image_folder=rootr, height=tarh, width=tarw, file_list = idlist[mi] )
这两种设置都只能生成横屏的视频，有没有大佬知道如何做能生成自适应原图横纵比的视频呀？

PS：pipe是这样的：
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.2-TI2V-5B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-TI2V-5B", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-TI2V-5B", origin_file_pattern="Wan2.2_VAE.pth", offload_device="cpu"),
],
redirect_common_files=False, tokenizer_config=ModelConfig(model_id="Wan-AI/Wan2.2-I2V-A14B", origin_file_pattern="google/*"),
)

Start by examining the VideoData height and width settings shown in the report, then trace how WanVideoPipeline handles those dimensions for the Wan2.2-TI2V-5B configuration. Confirm whether the current behavior forces landscape output and identify how original image aspect ratios are represented. Done means a documented or tested path for producing output that adapts to the source aspect ratio.