Fast CinemaDNG Processor on GPU

Updated: August 16, 2026

Fast CinemaDNG Processor decodes and develops CinemaDNG and RAW footage entirely on the GPU, faster than real time: it processes 4K CinemaDNG at 24–30 fps on NVIDIA GPUs through a full RAW-to-RGB pipeline with TIFF, EXR, and ProRes export.

How these numbers were measured. Fastvideo benchmarks are measured on NVIDIA GPUs, timing pure GPU processing at the resolution, bit depth, format and quality stated on this page. See the full benchmark methodology and test environment and the Fastvideo SDK benchmark report (PDF). Our tools are open source on GitHub and built on NVIDIA CUDA.

Check a single module. Every module in the Fastvideo SDK comes with a command-line sample. Run it on your own images on your own GPU and compare the result with the figures published here.

Check the whole pipeline. A module in the middle of a pipeline cannot be judged on its own: its input is hard to prepare, and its output is an intermediate product you cannot look at. What matters is how it changes the final image and the total time of the schema. Two of our applications run the complete pipeline and report the time of every stage — FastVCR (live cameras or RAW files from disk, so no camera is required) and Fast CinemaDNG (DNG, offline). The benchmark module is included in the free builds of both.

Detailed benchmark data. Measured results are published on our benchmark pages: all benchmarks.

We wrote these algorithms ourselves

Image processing has always been a trade-off between quality and speed. The methods that give a good result take minutes per frame; the fast ones leave halos and artifacts. The reason is the compute budget: on a CPU there was never enough of it, so algorithms were simplified — and much of the industry still runs on simplifications accepted decades ago.

Fastvideo’s codecs and ISP modules are neither ports of that legacy nor wrappers around third-party code. We designed — and in many cases invented — the algorithms behind our demosaicing, denoising, tone mapping and JPEG/JPEG2000 codecs, and wrote them for CUDA from the start, on top of image-processing expertise built since 2009. An algorithm moved to the GPU still carries the compromises it was simplified for; ours were never made under that constraint. On a modern GPU the budget is large enough that, for practical workloads, the choice between quality and speed no longer has to be made.

That is why you get a complete, tuned imaging solution rather than a bag of GPU primitives — and why, when your pipeline needs something specific, we can change the algorithm itself, not just call it.

Trusted by imaging teams

“Fastvideo SDK is lightning fast… we can now process a 28,800 image sequence set in under 10 minutes, instead of 10 hours.”
— Infinite Realities

Advantages of Fast CinemaDNG Processor

How fast is 4K CinemaDNG processing on GPU?

4K CinemaDNG processing at 24-30 fps, up to 60 fps
8K DNG photo editing with fast response
High performance DNG decoding on CPU
RAW Curves and Levels before demosaicing
High quality demosaicing with MG algorithm
High speed denoising for raw, luma and chroma
No proxies, 4K dailies can be processed in real-time
Smooth and scalable Player for CinemaDNG footages
MLV and DNG player, viewer, converter, transcoder
Software for footage Culling and Reviewing before editing

We have implemented full image processing pipeline on CUDA to make fast each stage of RAW data workflow. That gives us opportunity to significantly decrease data transfer time over PCI-Express bus and to improve total performance. Now we can do CUDA-based DNG rendering much faster than real time. At the same time, output image quality is comparable with Adobe Camera Raw (ACR) and LightRoom (LR), etc.

Fast CinemaDNG Processor GPU benchmark

In a RAW-to-RGB benchmark on an NVIDIA GeForce RTX 2080ti, Fast CinemaDNG Processor converted 3000 eight-bit 12-MPix DNG frames from an IO Industries camera per minute — 50 fps — through a full pipeline: read from SSD and upload to GPU, white balance, high-quality MG demosaicing, DCP profile, gamma, and JPEG (4:4:4, quality 100%) encoding back to SSD. Throughput was limited by the SSD at 600 MB/s, not by the GPU: raw GPU processing ran about 3 times faster, and GPU load stayed near 30% for a single-thread run, so a faster SSD or higher GPU load leaves significant headroom for RAW-to-JPEG acceleration.

See the full RAW-to-RGB benchmark test and discussion on LinkedIn.

Fast CinemaDNG Processor RAW-to-RGB benchmark	Value
GPU	NVIDIA GeForce RTX 2080ti
Input	8-bit 12-MPix DNG (IO Industries camera)
Throughput	3000 frames/min (50 fps)
Bottleneck	SSD at 600 MB/s (GPU ~3× faster)
GPU load	~30% (single-thread)

Features of CUDA RAW Processor

Import images: 8/10/12/14/16-bit DNG or CinemaDNG (compressed or uncompressed)
Other input image formats: MLV, SEQ, CR2, CR3, NEF, ARW, IIQ
Input video (CinemaDNG or MLV): 1080, 2K, 4K CDNG clips and more, up to 8K
Output formats for image series: JPEG (8/12-bit), TIFF (16-bit), EXR (16/32-bit)
Output video compression: Apple ProRes (MOV), Motion JPEG (AVI), any codec from external FFmpeg
Output Color Spaces: sRGB, Rec.709, AdobeRGB, Rec.2020, BMD Film, BMD Film 4K, BMD Film 4.6K
Highly optimized multi-threaded DNG and MLV reading, parsing and decoding on CPU
White Balance and Exposure Correction
Curves and Levels for each channel of RAW data
RAW denoising (float workflow)
Demosaicing algorithms: L7 (7×7), DFPD (11×11), MG (23×23)
Denoise adaptive algorithms for luma and chroma (float workflow)
Correlated Color Temperature and Tint control
Color correction with matrix or DCP profile
RGB and HSV Curves and Levels
3D LUTs for RGB and HSV, DCP support, Tone curve
Image rotation to arbitrary angle, other geometric transforms
LCP support: undistortion, chromatic aberrations, vignetting suppression
RGB Histograms for each frame in real-time
RGB Parade monitor
Resize algorithm: Lanczos (float workflow)
Adaptive local tone mapping
Sharpening with Unsharp Mask algorithm (float workflow)
Smooth playback of CinemaDNG, MLV, or SEQ sequences via OpenGL in real-time

Compatibility

Windows-10/11 (64-bit)
NVIDIA GPU driver 572.83 or later
CUDA-12.6 (64-bit)
NVIDIA GPU for laptop, desktop, server
Compatibility with FFmpeg to read/write MJPEG streams (FFmpeg is under LGPLv2.1)

CinemaDNG Workflow on CUDA

Which cameras and RAW formats does Fast CinemaDNG Processor support?

The software supports DNG/CinemaDNG files, produced either by DNG-enabled digital cameras or by Adobe software such as the Adobe DNG Converter or Adobe Lightroom (DNG 1.4 specification). Current list of supported cameras could be found at Adobe's list of supported cameras. Fast CinemaDNG Processor also supports 8/10/12/14/16-bit DNG created by 3rd-party applications.

DNG-enabled digital cameras

Blackmagic Design 4K Micro Studio Camera
Kinefinity cameras: KINEMINI 4K, KINEMAX 6K, KINERAW-MINI (Full HD), KINERAW-S35 (Full HD), MAVO, Terra 4K, Terra 6K
AJA CION 4K/UHD and 2K/HD Production Camera
DJI Zenmuse X5S, X5R or X7 cameras for Mavic, Phantom, Inspire, Matrice drones
Digital Bolex D16 and D16M
Atomos Ninja Shogun
Ikonoskop
XIMEA 4/5/8/12/20/26/50/65 MPix
IO Industries cameras: Volucam, Redwood, Flare, Victorem

Image/Video Processing Pipeline for DNG/CinemaDNG/MLV on CUDA

Ingestion of DNG images from HDD/SSD
- Multithreaded reading of RAW data from SSD to CPU memory
- CinemaDNG / MLV parsing
- DNG / CR2 / CR3 / NEF / ARW / IIQ parsing
- SEQ format support
- Multithreaded decoding for lossless JPEG
Data transfer from CPU to GPU
Data linearization and transform to 16-bit representation
Dark Frame Subtraction and Flat-Field Correction
Bad Pixel Correction
RAW histogram
White/Black points
White Balance (R, G1, G2, B)
Exposure compensation
Raw Curve for each channel of Bayer data
Raw denoising
Demosaicing
Adaptive noise suppression
Color Correlated Temperature and Tint control
Color Correction with matrix or DCP profile
Tone mapping
Histogram and RGB Parade for processed image
Monitor output
- Crop
- Resize to specified width and height in output window
- Sharp with USM
- Transform with monitor profile
- LUT 16-bit to 8/10-bit for each color channel
- Copy data to OpenGL texture
- Real-time monitor output via OpenGL (DNG video player)
- Histogram output for RAW and processed data
- RGB Parade output
Compressed video output
- Crop
- Resize to Full HD
- Sharp with USM
- LUT 16-bit to 8/10-bit for each color channel
- Native MJPEG encoding
- Native export to Apple ProRes 422 LT, 422, 422 HQ, 4444, 4444 XQ
- External compression via FFmpeg
- Data transfer from GPU to CPU memory
- Save compressed video to SSD/HDD
Image sequence output
- Data transform to 16-bit TIFF / EXR or to 8/12-bit JPEG for each frame
- Data transfer from GPU to CPU memory
- Save image sequence to SSD/HDD

Standard tasks to solve with Fast CinemaDNG Processor

Batch image processing for DNG image sequences
Real-time full frame rendering, compression, preview and playback for video streams with RAW data
Full image processing pipeline for industrial cameras with 8/10/12/14/16-bit Bayer frames, including SEQ support
Fast processing of MLV video files from Canon cameras with Magic Lantern firmware
RAW Converter: MLV to DNG, MLV to ProRes, CinemaDNG to ProRes, etc.
Converter from SEQ/MLV/DNG to ProRes

Workflow example for Fast CinemaDNG Processor

Upload video footage to computer (PC) from camera or card reader
Convert RAW files to DNG format with Adobe DNG Converter (option)
Create new project and import DNG images
DNG image processing with Fast CinemaDNG Processor
DNG video player
Export JPEG (8/12-bit), TIFF (16-bit), EXR (16/32-bit) files
Export video to Apple ProRes 422 LT, 422, 422 HQ, 4444, 4444 XQ
External FFmpeg options: export H.264/H.265 video in MP4, etc.

Roadmap for Fast CinemaDNG Processor

SEQ format support (RAW format from StreamPix software) - done
Support of IOI cameras: IOI format, transcoding of IOI files to CinemaDNG, IOI Player - done
Defringe module - done
Highlight Recovery on GPU - done
CUDA 12.6 support - done
Adaptive local tone mapping - done
Chromatic aberration removal - in progress
Noise profiles support - in progress
Linux (Ubuntu 22.04) support - in progress

About the author

Fyodor Serzhenko, PhD, is the founder and CEO of Fastvideo. He earned his PhD at the Moscow Institute of Physics and Technology (MIPT) in 1993. Since 2009 he has led the development of Fastvideo’s GPU-accelerated image codecs and ISP modules, including the Fastvideo SDK. Connect on LinkedIn.

Why you can trust these results

Fastvideo has built GPU-accelerated image processing software since 2009. Our codecs comply with the standards they implement — JPEG (ITU-T T.81 / ISO IEC 10918) and JPEG2000 (ITU-T T.800 / ISO IEC 15444) — and our tools are open source on GitHub. Every performance figure we publish is reproducible: see our benchmark methodology, download the Fastvideo SDK benchmark report (PDF), and measure it on your own GPU.