ComfyUI_save_image_sdli

ComfyUI_save_image_sdli
★ 1

SDLI保存潜变量(latents)VAE/TAESD
将隐空间(latents)与有损解码图像一起保存为SDLI文件，便于存档、重现与分享；支持可选VAE或TAESD回退并可对输出WebP缩放。
💡 将模型隐空间保存为含解码图像的SDLI文件，便于重现与分享。
🍴 1 Forks💻 Python🔄 2025-07-04
📦
网盘下载
复制链接后前往夸克网盘下载
https://pan.quark.cn/s/e00a65475347
📄 README

ComfyUI SaveImage SDLI

SDLI (Stable Diffusion Latents in Imagefile) is file format, that contains latents and lossy (decoded) image.

Detailed format is written at SDLI Tools.

Node: Save SDLI Image

Input:

  • samples: Latents
  • vae: vae (optional)
  • latent_type: Select proper SD type
  • reduction_ratio: shrink image for WebP. latents will not be shrinked.
  • filename_prefix: The prefix of output filename.
  • positive_prompt: Additional metadata.
  • negative_prompt: Additional metadata.

    • This node requires latents. (not image)

    If vae was not given, use TAESD.

    If you want to use TAESD, download models into models/vae_approx.

    Node: Preview SDLI Image

    Similar to Save SDLI Image.

    Generate SDLI (*.webp) for preview.

    You can download these webp via browser.