> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pulze.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Multichannel EXR

> Use a single EXR that holds all your render elements: LayerWise splits it into one layer per element.

If you save all render elements into one **multichannel EXR** instead of separate files, LayerWise handles it automatically. Pick one in either workflow and it's split into its layers, each treated as a separate render file.

No extra software needed, and it works the same on Windows and macOS.

## How it works

When you browse a folder or pick files, LayerWise reads the EXR's header and lists every layer inside it as its own file, named like a single-layer render: a Reflection layer in `cam01.exr` shows up as `cam01.Reflection.exr`. From there, matching, templates and updating work exactly as with separate files.

![A multichannel EXR's channel list and the render elements LayerWise matches from it](https://mintcdn.com/sourcelinekft/z3vWbLeBmT4OoYjW/images/layerwise/lw_exr_split.webp?fit=max\&auto=format\&n=z3vWbLeBmT4OoYjW\&q=85\&s=e5bea5f8a18b5a18d6bb66d55a1922c3)

One multichannel EXR becomes one render file per element

The layers are only extracted when you start the job. You'll see *Splitting cam01.exr* in the progress overlay, and you can cancel it. Each EXR is read once for all the layers you need from it.

## Layer rules

| Channels in the EXR | Becomes |
| - | - |
| `R`, `G`, `B` (and `A`) with no layer name | The beauty pass: **RGB\_color** for V-Ray, **Beauty** for Corona and FStorm |
| `A` with no layer name | **Alpha** |
| `Z` with no layer name | **ZDepth** |
| `Reflection.R`, `Reflection.G`, `Reflection.B` | A layer named `Reflection` |
| A layer with a single channel | That channel copied into grey |

* Only the beauty pass keeps its alpha channel. All other elements are written without alpha.
* **Cryptomatte** layers don't become layers. They feed [Cryptomatte masks](https://docs.pulze.io/layerwise/cryptomatte.md) instead.
* Corona **variance** (`Sqr`) layers are skipped.

## Where the split files go

* For **embedded** smart objects, the split layers are temporary files. Once placed, they live in the PSD.
* For **linked** smart objects, the PSD needs the files on disk. If you picked a **folder**, LayerWise writes them next to the EXR in a folder called `<name>_layerwise` and links them from there. Folder scans skip this folder, so nothing gets picked up twice.
* If you picked **individual files** and use linked smart objects, LayerWise can't write next to them, so the layers stay temporary and the job warns you. For linked layers, browse the folder instead.

## Supported EXRs

LayerWise splits **scanline** EXRs with **no compression, RLE, ZIPS, ZIP, PIZ or PXR24** compression. These are the defaults of V-Ray, Corona and FStorm.

EXRs it can't split get a note: **can't be split (multi-part, tiled, deep, or B44/DWA compression)**, and their layers can't be used as render elements. To fix this, re-save them from your renderer or frame buffer with a supported compression, or save the elements as separate files.

> [!TIP]
>
> Splitting a large EXR takes a few seconds (roughly 5 seconds for a 4K image with 14 layers). Linked output is written compressed and takes longer.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.