Fastvideo launches RawZipper GPU codec for 12-bit RAW Bayer compression
Fastvideo has released RawZipper, a GPU-accelerated codec designed to compress 12-bit RAW Bayer frames on NVIDIA and Jetson hardware. The software supports both lossless and lossy compression, allowing high-resolution imaging workflows to reduce data footprints while bypassing early RGB conversion.
Key Takeaways
- Supports both mathematically lossless and lossy (visually or nearly lossless) compression for 8-bit to 16-bit RAW Bayer or monochrome data.
- Achieves compression ratios ranging from 2:1 for lossless encoding to 20:1 for high-efficiency lossy applications.
- Runs natively on NVIDIA GPUs and Jetson hardware, maintaining high data rates for real-time capture and streaming.
- Maintains image fidelity for post-processing with peak signal-to-noise ratios (PSNR) typically exceeding 50-55 dB for 12-bit sensors.
- Eliminates the need for full RAW-to-RGB ISP conversion during initial recording, saving bandwidth and compute during live capture.
Why It Matters
RawZipper addresses a critical bottleneck in high-resolution production and industrial imaging: the massive data volume of uncompressed sensor output. By enabling high-ratio RAW compression directly on GPUs, Fastvideo allows operators to defer resource-heavy demosaicing to post-production without sacrificing sensor-level detail. For the broader ecosystem, this shifts the pressure from transmission bandwidth to edge-computing efficiency, particularly on mobile platforms like Jetson. This development signals a move away from general-purpose video codecs like HEVC or AV1, which lack native support for 12-bit RAW Bayer domains. Watch for integration into autonomous vehicle data-logging stacks where multi-camera 12-bit RAW feeds currently overwhelm standard storage architectures.
Additional Context
The launch of RawZipper arrives as industry reliance on GPU-accelerated video infrastructure hits record levels. Per a July 2026 report by Netint, GPUs now command a 72% adoption rate in hardware-accelerated streaming and production environments, largely powered by NVIDIA’s NVENC ecosystem. However, that dominance faces mounting friction from rising power consumption and the inability of standard hardware encoders to handle the specific bit-depth and pattern requirements of RAW Bayer data. While traditional interframe codecs like HEVC and AV1 have seen widespread adoption for distribution, they remain poorly suited for the high-fidelity demands of scientific and cinematic RAW workflows. Competitive pressures are also shifting toward edge performance. In October 2025, NVIDIA highlighted that Level 4 autonomous driving systems generate petabytes of data annually, necessitating new 'cloud-to-car' architectures that prioritize intelligent data reduction at the sensor level. Simultaneously, hardware benchmarks from Puget Systems in September 2025 indicated that while NVIDIA’s latest RTX architectures continue to lead in RAW media processing, performance gains are increasingly tied to specific software-optimization layers rather than raw silicon improvements. By targeting the Bayer domain specifically, Fastvideo is filling a gap that general-purpose hardware refreshes have historically overlooked.
Read full article at fastcompression.com
Get this in your inbox → Subscribe
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source