About the usage of diffusers for only denoising effect

[momo1986]

Hi guys,

I want to use diffusers for the denoising effect.

My code is:

from diffusers import StableDiffusionImg2ImgPipeline
diffusion_pipeline = StableDiffusionImg2ImgPipeline.from_pretrained("nitrosocke/Ghibli-Diffusion", torch_dtype=torch.float32, use_safetensors=False, safety_checker=None, local_files_only=False, from_tf=False).to(device)
noisy_image_pil = Image.fromarray(np.uint8(img))
denoised_image = diffusion_pipeline(
         prompt="high-quality clean image",  # 或留空 ""
         image=noisy_image_pil,
         strength=0.4,
        guidance_scale=2.0,
        ).images[0]

However, the denoising effect is unsatisfactory. How can I resolve the issue?

My requirement is to remove noise without changing the semantic context of the original image. How do I call the API?

Thanks & Regards!
Momo