TAU-HOME.COM
LOADING

Veyra: Real-Time GPU Video Enhancement and Frame Generation for Windows

An overview of Veyra, an open-source C++ and DirectX 12 utility that processes local video files, capture cards, and game console streaming through a real-time

tau · November 15, 2026

#Veyra #GPU-Upscaling #Super-Resolution #Frame-Generation #DirectX12 #Windows-Tools

Veyra: Real-Time GPU Video Enhancement and Frame Generation for Windows

Enhancing video quality or smoothing out frame rates on personal computers typically requires lengthy offline rendering workflows. The open-source project Veyra changes this equation by leveraging modern C++ and native DirectX 12 graphics pipelines to deliver real-time post-processing directly to video streams and live capture feeds.

Veyra video enhancer professional interface showing node-based real-time video processing pipeline

Image source: https://github.com/Likely7/Veyra-NRVideo

Published on GitHub under the GNU General Public License v3.0 (GPL-3.0), Veyra is a modular real-time video processor designed for Windows 11. It accepts local video files, hardware HDMI capture cards, PlayStation 5 and Xbox console streams, Moonlight and Sunshine PC remote streaming inputs, and standard desktop captures. By applying real-time super resolution upscaling, multi-layered noise filtering, and frame generation on the fly, the project has quickly surpassed 79,000 portable package downloads across community releases.

Real-Time Post-Processing Architecture and Node-Based Editing

Traditional video enhancement software processes footage frame-by-frame before saving the result into a new video file on disk. In contrast, Veyra is engineered as a zero-latency or near-zero-latency playback engine that intercepts incoming frames, submits them through GPU computing passes via DirectX 12, and renders the result to the display in real time.

The application interface is structured into dedicated operational modes: Home and Cinema for streamlined playback, alongside Professional List and Nodes view for granular parameter adjustments. In the Nodes workspace, users can visually connect incoming video feeds to sequential processing blocks. Each node operates with discrete settings, and complete routing configurations can be saved as persistent sessions and reusable presets.

A standout aspect of this pipeline is its layered Noise Reduction (NR) framework. Rather than forcing a single global denoising slider that often blurs fine textures, Veyra allows users to configure independent NR layers with customizable scale factors ranging from 25% to 100% of the native inference resolution. The software features nine granular manual adjustment parameters—including hue, chroma, neutral tint, source color retention, lightness preservation, shadows, highlights, local compression, and temporal stability. These controls help eliminate noise artifacts while safeguarding facial features and contrast edges.

Dual Super Resolution (SR), Frame Generation (FG), and Anti-Aliasing

Veyra integrates multiple real-time rendering algorithms to elevate standard-definition or compressed console inputs to high-clarity output:

  • Dual Super Resolution (SR): The engine supports manufacturer-specific neural upscaling libraries, including NVIDIA RTX Video SR, DLSS, AMD FidelityFX Super Resolution (FSR), and Intel XeSS options. Users can configure a dual-pass pipeline that runs initial upscaling followed by controlled downsampling or secondary model passes, yielding crisper edges without severe haloing.
  • Real-Time Frame Generation (FG): Veyra deploys sparse-noise-tolerant full-frame deduplication and smoothed motion anchors to calculate intermediate frames dynamically. By capping the adaptive vertical synchronization (VSync) wait time at 3 milliseconds, the pipeline minimizes stutter and maintains fluid motion even when the incoming feed fluctuates.
  • Anti-Aliasing (AA): Dedicated FXAA, SMAA, and MLAA shaders can be inserted into the post-processing chain. Benchmarks on an RTX 5070 running a standard 4K output demonstrate minimal execution overhead: approximately 0.15 ms for FXAA and around 0.30 ms for SMAA and MLAA.
  • Low Latency Control: Integration with NVIDIA Low Latency technologies (Off / On / On + Boost) ensures that input-to-display latency is held to a minimum, which is critical when playing responsive console games over capture cards or local network streaming.

The pipeline also incorporates 468-point contour face enhancement with original-face restoration, HDR wide-gamut reconstruction, and an export queue capable of recording the processed visual stream alongside synchronized audio directly into MP4 or MKV containers.

System Prerequisites and Deployment Considerations

Because Veyra executes demanding shader models and machine learning inferencing concurrently on the graphics adapter, hardware specifications are an essential consideration.

The application strictly targets 64-bit Windows 11 running DirectX 12. Stacking multiple layers of high-resolution noise reduction alongside dual super resolution and frame generation places substantial demands on video memory (VRAM). While the project lists modern GPUs such as the RTX 5070 in its testing configurations, users on less powerful hardware must manage performance by throttling individual NR inference sizes down to 50% or turning off non-critical filters.

Veyra is distributed as self-contained portable archive files that run without standard installation overhead. However, binaries are split into vendor-specific packages (such as Veyra-2.1.5-NVIDIA-win64-portable.zip and dedicated AMD packages). Users should verify their graphics hardware vendor prior to downloading to ensure full compatibility with the underlying hardware acceleration libraries.

Sources

Impact-Site-Verification: 9b3ebd10-6c45-44cf-a014-5b8c19b7b285