A DLSS-G rival whose repo is mostly a tape recorder

What it does

Apex Neural Rendering is a suite of neural rendering features: frame generation (ANR FG, in beta), plus super resolution and ray reconstruction denoising, both listed as “beta soon.” The frame generator installs into games via OptiScaler and works across NVIDIA, AMD, and Intel GPUs on Windows and Linux/Proton, including the Steam Deck. Crucially, the ANR source itself is “coming soon” — what’s actually in this repository is nr-record, a capture-and-compare toolkit.

The interesting bit

nr-record records the exact inputs a game hands its frame generator, losslessly, then replays them through any frame-generation DLL you own — ANR FG or the game’s own DLSS-G — and produces synced side-by-side videos with pixel zoom and comparison marks. It’s a benchmark rig for upscalers, and it’s the rare vendor that ships the measuring tape alongside the product.

Key highlights

• Frame generation across RTX 20–50, GTX 16, Radeon RX 5000–9000, Intel Arc A/B, and Steam Deck
• Works on Windows 10/11 and Linux/SteamOS via Proton
• nr-record captures DLSS-G inputs losslessly (~26 MB per 4K frame) and replays them through any generator DLL
• Comparison viewer with synced playback, divider, side-by-side, pixel zoom, and video/image export
• Can also record upscaler inputs (DLSS, FSR, XeSS) and replay those; built with Bazel in C++

Caveats

• The ANR source code is not in the repo — only nr-record is; the core product is closed binaries for now
• ANR SR and RR are announced but not yet available
• DLSS-G replays require an RTX 40/50 GPU, and the tooling leans on OptiScaler (GPL-3.0) as its injection mechanism

Verdict

Worth watching if you’re on AMD or Intel and want DLSS-G-class frame generation, or if you enjoy rigorously comparing upscalers. If you wanted to read the neural rendering code itself, the repo will disappoint you — for now.