Skip to content

HDR Merge

Give it a bracket of the same scene, such as dark, normal and bright. It returns one photo with the best-exposed parts of each shot.

HDR Merge — three exposures and the merged result
In ComfyUI Add Node → ComfyUI-JH-PixelPro → color → Color: HDR Merge (bracket)
Good for Interiors with bright windows, landscapes with a bright sky, real estate, any high-contrast scene shot as a bracket
Speed About 1 s for three 2 MP frames, on the CPU

When to use it

A single exposure often cannot hold both a bright window and a dark room. A bracket (for example -2, 0 and +2 EV) captures both. This node picks, pixel by pixel, the well-exposed, contrasty and colourful parts of each frame and blends them smoothly.

The method is exposure fusion. It needs no exposure values and no camera response curve, and the result is ready to view. There is no tone-mapping step with sliders to tune.

The result is a normal display-ready image, not a 32-bit HDR file. Treat it as a starting point and finish it with Auto Tone or ColorLab.

Quick start

  1. Load the bracketed photos and combine them into one batch with Batch Images (or Load Image Batch). The order does not matter, and all frames must be the same size.
  2. Connect the batch to Color: HDR Merge (bracket) and add a Preview Image.
  3. Keep align and deghost on.
  4. Finish with Auto Tone or ColorLab if you want to adjust the look.

Settings

Setting Default What it does Tip
images A batch of 2 or more bracketed shots, in any order, all the same size. Use Batch Images or Load Image Batch to build the batch.
align on Lines up hand-held shots by fixing small shifts (translation only). Turn off for shots taken on a tripod, if you want the fastest result.
deghost on Where something moved between shots (people, leaves), uses the middle exposure, brightness-matched, so nothing appears twice. Turn off for fully static scenes.

Output

Output What it is
image One merged image.

Tips from retouching

  • Shoot -2, 0 and +2 EV on a tripod, or use a fast burst, so the frames match as closely as possible.
  • Big movements or rotation between shots are not aligned. Reshoot, or fix the rotation first with Auto Align if the frames allow it.
  • Moving subjects such as people, cars and tree branches are handled by deghost. Check them at 100 % after merging.
  • The merge looks natural and slightly flat. Add contrast and colour afterwards with Auto Tone, ColorLab or Shadows/Highlights Pro.
  • Two frames also work, but three or more give smoother transitions.
Under the hood

Frames are sorted by brightness and the middle one is the reference. With align, each frame is shifted onto the reference with median-threshold-bitmap alignment (OpenCV AlignMTB). A shift larger than 5 % of the long edge is treated as a failed alignment and ignored.

With deghost, each frame is first scaled to the reference brightness. Where it still disagrees strongly, which means movement, the frame takes the brightness-matched reference pixels with a soft edge.

The frames are fused with Mertens exposure fusion (Mertens, Kautz and Van Reeth, 2007): a multi-resolution blend weighted by contrast, saturation and well-exposedness. It runs on the CPU in OpenCV. Source: core/hdr_merge.py, wrapper nodes/hdr_merge_node.py.

Example workflow

HDR Merge example workflow turns one cafe photo into three exposures (-2, 0 and +2) with ColorLab, combines them with Batch Images, and merges them. Copy workflows/sample_cafe.jpg into ComfyUI/input/ (Pexels License).