Open-Source DLSS 5 Visual Enhancer Brings NVIDIA's Neural Rendering to Any Video
A community Windows app wraps NVIDIA's DLSS 5 neural rendering pipeline into a local Gradio UI for upscaling images and video.
- DLSS 5 Visual Enhancer wraps NVIDIA's neural rendering into a local Gradio app for images and video.
- Supports batch image jobs plus H.264, HEVC, AV1, and ProRes Proxy video output up to 8K.
- Uses OpenCV DIS optical flow to substitute for game-engine motion vectors on video frames.
- Requires RTX 40 or 50 primarily, with RTX 30 as a slower beta path; RTX 20 is rejected.
- Repo ships only the Python glue; NVIDIA DLLs, FFmpeg, and runtime files must be sourced separately.
- Pipeline is 8-bit SDR only, so HDR masters are downconverted rather than preserved.
DLSS 5 Visual Enhancer brings NVIDIA’s neural rendering to local media
An open-source GitHub project called DLSS 5 Visual Enhancer wraps NVIDIA’s DLSS 5 neural-rendering path in a local browser interface for processing images and videos. The Windows application decodes media, constructs inputs that DLSS normally receives from a game engine, and sends frames to a Direct3D 12 worker.
The repository has collected 1 star and contains the Python interface and orchestration code. A working renderer also requires omitted executables, DLLs, an embedded Python environment, and proprietary runtime files. The project is unaffiliated with NVIDIA, ReShade, RenoDX, and FFmpeg.
| Area | Current scope |
|---|---|
| Interface | Local Gradio web UI |
| Platform | 64-bit Windows with Direct3D 12 |
| Primary GPUs | GeForce RTX 40 and 50 series |
| Media pipeline | 8-bit SDR RGBA |
| Repository status | UI and orchestration code without a complete native renderer |
DLSS outside a game engine
Games normally provide DLSS with motion vectors, depth buffers, exposure data, and masks generated during rendering. Visual Enhancer approximates that contract for ordinary media. It decodes each input to 8-bit SDR RGBA, normalizes orientation and dimensions, and selects a render size based on the chosen DLSS mode.
Still images use zero motion vectors and reset temporal history for each file. Video processing uses OpenCV’s DIS optical-flow algorithm to estimate current-to-previous motion between frames. The pipeline resets history on the first frame and after a detected scene change.
A Direct3D 12 worker streams the prepared frames through the
This story is for Pro members
You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.
Comments
