Fastvideo: high speed imaging and GPU software
Updated:
Modern cameras output data faster than a CPU can process it at full quality. That is exactly what we solve — the whole image pipeline on an NVIDIA GPU, from RAW to compressed output, since 2009.
- JPEG up to 5500 FPS on CUDA
- 10–30× faster than CPU
- Jetson to server GPUs
- GPU imaging since 2009
How to work with us
Fastvideo SDK
The full GPU pipeline and codecs in one C++/CUDA library — Jetson to server GPUs, RAW to compressed video. For teams integrating imaging into their own application.
Open the SDK →Fastvideo applications
Working applications, not code: FastVCR capture, ALTMapper tone mapping and dehaze, JPEG/JPEG2000 codecs, debayer and RawZipper.
See the software →High-speed cameras
Cameras on CMOS sensors with Camera Link and PCI Express — designed and built in-house, for scientific and industrial work.
See the cameras →Systems & custom development
High-speed imaging systems, hardware/software complexes and custom development tuned to your sensors and workflow.
See the projects →Fastvideo is a privately held company based in Dubna, Moscow region, specialising in high speed imaging systems and high performance GPU‑based software. We design hardware, firmware and software for high speed imaging, and we develop GPU image and video processing that reaches performance conventional CPU code cannot.
- Company: privately held, based in Dubna, Moscow region, Russia
- What we build: high-speed cameras and imaging systems — hardware, firmware and software in one team
- GPU software: JPEG and JPEG2000 codecs, demosaicing, denoising, tone mapping and a full ISP SDK on NVIDIA CUDA
- Applications: machine vision, high-speed imaging, digital cinema, streaming and broadcasting
What we do
Our team of image‑processing and parallel‑computing engineers builds solutions for image compression and decompression, demosaicing, denoising, tone mapping, colour correction, resizing, sharpening, and video encoding/decoding. These are used across high speed imaging, machine vision, streaming, digital cinema, 3D/VR and broadcasting. Alongside the software, we design and manufacture high speed cameras for scientific and industrial applications.
Technologies and experience
- High speed CMOS sensors and interfaces (Camera Link, PCI‑Express)
- High performance NVIDIA CUDA computations for image and video processing
- GPU and CPU programming, algorithm design and low‑level optimisation
- Software for high performance camera applications and custom GPGPU solutions
Flagship GPU software
| Product | What it does |
|---|---|
| CUDA JPEG codec | The fastest Baseline JPEG on GPU |
| CUDA JPEG2000 codec | Real-time J2K encode/decode on GPU |
| Debayer on CUDA | High quality demosaicing (HQLI, L7, DFPD, MG) |
| GPU denoiser | Real-time Bilateral and NLM denoising |
| Tone mapper on GPU | Single-shot HDR with ALTMapper |
| Image & Video Processing SDK | Full GPU ISP pipeline on CUDA |
See the full list on the products page, our current projects, and the low‑latency software for machine vision cameras.
How we work
You can watch our software solve your task, on your own data — before any commitment.
Evaluate
Free No cost, no commitmentTell us your task
A short note on what you are building and the problem to solve — no full specification needed, we shape it together.
Get the demo
We send a demo SDK or a ready application for your platform (Windows, Linux or Jetson), the manual and a couple of clarifying questions.
See it on your own data
Send us your images — we process them, show the result and go through it together on a call. It is free: you confirm the result on your own material before any spend. An NDA is available on request; many customers sign one at this stage, and we are glad to.
Rollout
By contract Price known up frontWe propose the path
Depending on the task: the SDK to integrate yourself, a ready application, or a one-time setup tuned to your sensors and workflow. Deeper work is always fixed-price, never hourly — you know the cost up front.
Simple ordering
We send a quotation, you send an order (contract/PO), we issue the invoice. Licence terms ship with the software — nothing extra to negotiate, you can start right away.
Support and updates
Ongoing support and new releases keep you on the current version.
Straight from the first email. You take on no commitment until you have seen the result on your own data — and after that, ordering is quick and the terms are already in your hands.
About the author
Fyodor Serzhenko, PhD — founder and CEO of Fastvideo. He graduated from the Moscow Institute of Physics and Technology (MIPT) and earned his PhD in 1993. Since 2009 he has led the development of Fastvideo’s GPU codecs and ISP modules, including the Fastvideo SDK. LinkedIn.
Why you can trust Fastvideo
Fastvideo has built GPU-accelerated imaging 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 part of our tooling is open source on GitHub, so the claims can be checked in code.
Frequently asked questions
What does Fastvideo do?
Fastvideo develops and manufactures high-speed cameras and imaging systems, and builds GPU-accelerated image and video processing software on NVIDIA CUDA — covering hardware, firmware and software in one team.
Where is Fastvideo based?
In Dubna, Moscow region, Russia. Fastvideo is a privately held company; contact details are on the contacts page.
Does Fastvideo make cameras?
Yes. Alongside its GPU software, Fastvideo designs and manufactures high-speed cameras with high-speed CMOS sensors and Camera Link / PCI Express interfaces, for scientific and industrial applications.
What GPU software does Fastvideo offer?
JPEG and JPEG2000 codecs, demosaicing, denoising, tone mapping, resizing and video encoding/decoding on NVIDIA CUDA, plus a full Image & Video Processing SDK.