EXR vs HDR
Radiance HDR stores radiometric RGBE or XYZE values compactly using a shared exponent. It suits lighting environments, but cannot preserve EXR-style arbitrary channels or alpha, and the shared exponent limits precision.
Image format
HDR raster data for rendering and compositing, with named channels and lossless or lossy codecs.
Explore the format
Choose a target format for your EXR images.
Marked paths may need steps such as flattening, rasterization, tone mapping or animation loss.
Choose a source format to convert to EXR.
Marked paths may need steps such as flattening, rasterization, tone mapping or animation loss.
OpenEXR
Industrial Light & Magic developed OpenEXR for high-dynamic-range image interchange in visual effects. Named channels store image or processing data rather than requiring one RGB layout. Samples can be 16-bit floating-point half, 32-bit float or 32-bit unsigned integers. This preserves a working range beyond ordinary display-encoded photo files.
EXR supports scanlines or tiles, optional alpha, typed header attributes and several compression choices. Multipart structures hold related images, and deep data stores multiple samples at different depths per pixel. These are not ordinary Photoshop editing layers or a timed animation; readers may support only a subset of the structures.
Use EXR for render passes, scene-linear compositing and HDR working images. Keep channel names, colour-space assumptions and premultiplication explicit. A visible preview needs a display transform or tone mapping. Exporting that preview to a simple image can discard extra channels, deep structure and the original working dynamic range.
EXR supports lossy and lossless codecs. Choose one for the sample types and required precision; .exr alone does not guarantee exact channel preservation.
OpenEXR grows from an ILM production format into an open project with contributions from multiple studios.
Different formats, different trade-offs. Choose for the image and the workflow.
Radiance HDR stores radiometric RGBE or XYZE values compactly using a shared exponent. It suits lighting environments, but cannot preserve EXR-style arbitrary channels or alpha, and the shared exponent limits precision.
TIFF provides tagged raster storage with multiple images and several sample types and compression schemes. It is a useful interchange choice for scanning and high-precision editing, but the receiving tool must support the particular TIFF variant.
PSD preserves Photoshop layers, masks and document resources. Keep it for continued editing; a flattened raster export is easier to distribute but cannot recreate the original layer structure.
AVIF uses AV1 coding and supports alpha, image sequences and higher bit depths for HDR workflows. It is worth evaluating for final delivery; codec settings and the receiving application determine whether its additional features survive.
Industrial Light & Magic developed OpenEXR for high-dynamic-range image interchange in visual effects. Named channels store image or processing data rather than requiring one RGB layout. Samples can be 16-bit floating-point half, 32-bit float or 32-bit unsigned integers. This preserves a working range beyond ordinary display-encoded photo files.
EXR supports lossy and lossless codecs. Choose one for the sample types and required precision; .exr alone does not guarantee exact channel preservation.
Channels are named sample arrays. Multipart and deep structures add images or depth samples, but do not automatically preserve Photoshop masks, blend modes or editable text.
Check channel types, compression, tiles and multipart/deep support. Agree on colour space and alpha; a reader displaying RGB may omit essential processing channels.
SDR export needs tone mapping and channel selection. RGBE cannot retain alpha, negative values or arbitrary channels. PNG/JPEG previews cannot reconstruct deep or multipart data.
Use EXR for render passes, scene-linear compositing and HDR working images. Keep channel names, colour-space assumptions and premultiplication explicit. A visible preview needs a display transform or tone mapping. Exporting that preview to a simple image can discard extra channels, deep structure and the original working dynamic range.