[Bug]: The Picture Generated through foreground_to_blend is blur and dark

[Ning570]

What happened?

I am a newbie and I'm facing some bug that I cannot find the solution.
When I use the foreground_to_blend workflow to generate a car poster, the output_picture is very dark and has a mosaic effect.
Even more strangely, if I used the foreground_image generated by LayerDiffuse to generate a new blend_image, such a result would not occurred.

Steps to reproduce the problem

my workflow:

the Prompt and CheckPoint I used:

Output:

What should have happened?

How can I get a better_quality output?I would be extremely grateful if you could help me

Commit where the problem happens

ComfyUI:
ComfyUI-layerdiffuse:

Sysinfo

Graphics Card：Nvidia
OS:Intel

Console logs

Requested to load SDXL
Loading 1 new model
WARNING SHAPE MISMATCH diffusion_model.input_blocks.0.0.weight WEIGHT NOT MERGED torch.Size([320, 8, 3, 3]) != torch.Size([320, 4, 3, 3])
Merged with diffusion_model.input_blocks.0.0.weight channel changed from torch.Size([320, 4, 3, 3]) to [320, 8, 3, 3]

Workflow json file

fg2ble.json

Additional information

No response