ComfyUI-HQ-Image-Save

ComfyUI-HQ-Image-Save
★ 99

高精度保存EXR 32位TIFF 16位色彩保真
为 ComfyUI 提供高精度图像保存节点，支持 TIFF 16 位和 EXR 32 位，保留浮点色彩与潜变量，适用于应用 LUT 或色彩校正时最大限度保留色彩精度。
💡 在应用 LUT/色彩校正后保存高保真图像或潜变量文件。
🍴 15 Forks💻 Python🔄 2026-03-09
📦
网盘下载
复制链接后前往夸克网盘下载
https://pan.quark.cn/s/b45acaa3411d
📦 requirements.txt
imageio
opencv-python
comparison
📄 README

ComfyUI-HQ-Image-Save

Nodes:

  • Image
  • Load EXR (Individual file, or batch from folder, with cap/skip/nth controls in the same pattern as VHS load nodes)
  • Load EXR Frames (frame sequence with start/end frames, %04d frame formatting for filenames)
  • Save EXR (RGB or RGBA 32bpc EXR, with full support for batches and either relative paths in the output folder, or absolute paths with version and frame number formatting, and overwrite protection)
  • Save EXR Frames (frame sequence with %04d formatting, optionally save workflow as GUI .json and/or API .json)
  • Save Tiff (RGB 16bpc TIFF, outdated)
  • Latent
  • Load Latent EXR (Same VHS style controls now)
  • Save Latent EXR (4 channel latent -> RGBA 32bpc EXR)

    • Overview

    Save and load images and latents as 32bit EXRs

    I recommend adding the --fp32-vae CLI argument for more accurate decoding. If you get an error saying that the OpenEXR codec is disabled, see this issue.

    Scatterplot of raw red/green values, left=PNG, right=EXR. PNG quantizes the image to 256 possible values per channel (2^8), while the EXR has full floating point precision.

    For latent EXR viewing purposes, if you want a cheap approximation of RGB values from the four latent channels, use this formula:

    r = (0.298 * r + 0.187 * g - 0.187 * b - 0.184 * a) * 0.18215
g = (0.207 * r + 0.286 * g + 0.189 * b - 0.271 * a) * 0.18215
b = (0.208 * r + 0.173 * g + 0.264 * b - 0.473 * a) * 0.18215

    Known Issues

  • No load TIFF node, and the TIFF save is bad/outdated