Skip to content

System Requirements

Visual Enhancer runs on 64-bit Windows (Windows 10 or Windows 11) and scales dynamically across NVIDIA RTX, DirectX 12, CUDA / AMD ROCm / Intel XPU, Vulkan, and CPU execution paths.

Feature / Stage Required Hardware / Runtime Notes & Recommendations
DLSS Neural Rendering, DLSS Super Resolution, DLSS Ray Reconstruction NVIDIA RTX GPU (compute capability >= 7.5, Turing RTX 20 series or newer) Uses NVIDIA NGX on CUDA; NVIDIA Display Driver up to date
DLSS Frame Generation NVIDIA RTX GPU (compute capability >= 7.5) RTX 40 / 50 series recommended; Hardware-Accelerated GPU Scheduling (HAGS) recommended in Windows Settings for optical-flow frame generation
RTX Video Super Resolution & RTX Video HDR NVIDIA RTX GPU (compute capability >= 7.5) Runs through native RTX Video SDK acceleration
NVIDIA VFX (Video Effects) NVIDIA RTX GPU (compute capability >= 7.5) Video-only stage
AMD FSR 3 Frame Generation Any DirectX 12 GPU (NVIDIA, AMD, Intel) Enumerated by DXGI adapter in Settings > AI Processing GPU (DirectX 12 / FidelityFX)
ReShade Effects DirectX 12 GPU (NVIDIA, AMD, Intel) Uses AMD FidelityFX Optical Flow when temporal stability is enabled; NVIDIA GPUs additionally accelerate supported video paths via CUDA–D3D12 interop
MoGe-2 AI Depth (ReShade depth) DirectX 12 GPU Powered by ONNX Runtime DirectML (moge-2-vitb)
ClearMagnetar, FocusPulsar, VisionNext, RazorNext, PeakMagnetar, OptiPulsar GPU supported by NeuroFrame Runtime Choose Auto, CUDA (NVIDIA), XPU (Intel), or ROCm / HIP (AMD) in Settings > Upscale > AI Inferencing Framework
Scaling, Sharpening, Grain, Coloring, Style Transfer, Compare Vulkan adapter / DirectX 12 / CPU Video filter stages prefer Vulkan hardware filters when selected under Settings > Decoding / Encoding Device; Style Transfer & Compare run smoothly across DXGI / CPU
Format Conversion / Re-encoding (zero active pipeline cards) CPU only (no RTX GPU required) When no pipeline stages are enabled, Render performs a clean decode-to-encode pass-through conversion
  • OS: Windows 10 or Windows 11 (64-bit x64).
  • Hardware-Accelerated GPU Scheduling (HAGS): Enable in Windows Settings > System > Display > Graphics > Default graphics settings when running DLSS Frame Generation or RTX Video pipelines on RTX 40 / 50 series GPUs.
  • NVIDIA NVENC Hardware Encoding: Supported on NVIDIA GPUs for H.264 (NVIDIA NVENC), H.265 (NVIDIA NVENC), and AV1 (NVIDIA NVENC). Non-NVIDIA systems automatically use the bundled software encoders (H.264, H.265, AV1, VP9, ProRes Proxy, ProRes HQ, FFV1 Lossless RGB 10-bit).

Visual Enhancer validates input and output dimensions before opening GPU sessions or encoders:

Scope Minimum Maximum Alignment / Constraint
General Image & Video Pipeline 1 × 1 px (image), 2 × 2 px (video) 32,768 × 32,768 px Decoder memory budget capped at 32,768 MiB per image
DLSS Neural Rendering 64 × 64 px 16,384 × 16,384 px DLSS native texture limit (16,384 px per side)
DLSS Frame Generation 64 × 64 px (scaled min) 16,384 × 16,384 px DLSS texture limit (16,384 px per side)
AMD FSR 3 Frame Generation 64 × 64 px 16,384 × 16,384 px 64 px minimum per side, 16,384 px maximum per side
ReShade Stage 16 × 16 px Pipeline maximum 16 px minimum per side
WebP Export 1 × 1 px 16,383 × 16,383 px Libwebp container cap
H.264 (OpenH264 / NVENC) Even dimensions 4,096 × 4,096 px OpenH264 software encoder also caps frames at 36,864 macroblocks
H.265 (NVIDIA NVENC) & AV1 (NVIDIA NVENC) Even dimensions 8,192 × 8,192 px 10-bit p010le when Preserve / Export 10-bit HDR is enabled
H.265 (Kvazaar software) Multiple of 8 on both axes 32,768 px 8-bit SDR only; both width and height must be divisible by 8
AV1 (SVT-AV1 software) Even dimensions 16,384 × 8,704 px 10-bit yuv420p10le when HDR is enabled
VP9, ProRes Proxy, ProRes HQ, FFV1 2 × 2 px 32,768 × 32,768 px Exempt from the even-dimension rule; all support 10-bit HDR