Skip to content

Pipeline Overview

Visual Enhancer builds your processing graph as a vertical stack of stage cards in the Pipeline rail. Enabled cards execute strictly from top to bottom, passing each stage’s output directly into the next active stage.

All stages start disabled until you toggle them on. Image mode exposes 12 stages, while Video mode expands the stack to 18 stages with HDR conversion, frame generation, and NVIDIA VFX video cards.

Order Image Pipeline (12 Stages) Video Pipeline (18 Stages) Focus Area
1 Denoising Denoising Cleanup
2 Deblurring Deblurring Cleanup
3 DLSS Neural Rendering DLSS Neural Rendering DLSS Neural Rendering
4 Scaling Scaling Upscaling
5 DLSS Super Resolution DLSS Super Resolution Upscaling
6 DLSS Ray Reconstruction DLSS Ray Reconstruction Ray Reconstruction & Motion
7 RTX Video Super Resolution RTX Video Super Resolution Upscaling
8 Upscale Upscale Upscaling
9 — RTX Video HDR Color, HDR & ReShade
10 — DLSS Frame Generation Ray Reconstruction & Motion
11 — FSR Frame Generation Ray Reconstruction & Motion
12 ReShade ReShade Color, HDR & ReShade
13 Coloring Coloring Color, HDR & ReShade
14 Sharpening Sharpening Finishing
15 Grain Grain Finishing
16 — Video Super Resolution Ray Reconstruction & Motion
17 — Upscale (Lightweight) Ray Reconstruction & Motion
18 — Video Frame Generation Ray Reconstruction & Motion
  • Enable or bypass: Toggle the switch on the right side of any card to include or skip that stage without losing its parameter values.
  • Inspect controls: Click the card header or chevron to expand its settings. Deblurring and Image-mode DLSS Ray Reconstruction contain no extra controls and remain single-row toggles.
  • Reorder execution: Drag any stage card by its title bar, or use the card’s up/down buttons, to change when it runs relative to other stages (for example, placing Scaling after an AI upscaler to downsample cleanly, or moving Coloring before Grain).
  • Reset defaults: Click Reset neural rendering settings at the bottom of the panel to restore default stage order and parameters for the active workflow while keeping your GPU device, container, and codec choices intact.
Visual Enhancer Pipeline stage cards showing toggles, reorder chevrons, Denoising controls, and footer Preview and Render actions
The Image pipeline rail showing stage switches, expand/reorder chevrons, footer Preview/Reset controls, and Render Image(s).

RTX Video HDR converts standard dynamic range (SDR) video into 10-bit PQ HDR (BT.2020 / SMPTE ST 2084). Because several enhancement stages only process SDR frames, you must place RTX Video HDR after any active SDR-only stage:

  • Cleanup: Denoising, Deblurring
  • Video Upscaling: Video Scaling, Video DLSS Super Resolution, Video RTX Video Super Resolution, Video Upscale (PeakMagnetar, OptiPulsar, RazorNext, VisionNext), Upscale (Lightweight)
  • Finishing: Sharpening whenever Sharpness is above 0%
Stage Group Rule
DLSS Super Resolution vs DLSS Ray Reconstruction Mutually exclusive — only one may be enabled in a run.
Frame Generators (DLSS, FSR, Video Frame Generation) Only one frame-generation stage may be enabled per workflow; switching one on turns off any other active frame generator.
NVIDIA VFX Chaining Adjacent Video Super Resolution, Upscale (Lightweight), and Video Frame Generation stages share CUDA GPU buffers without host-memory copies. Upscale (Lightweight) requires 8-bit SDR input and cannot directly follow spatial Video Super Resolution (Standard Upscaling Modes or High-Bitrate Modes).
RTX VSR + VFX Combination When RTX Video Super Resolution runs alongside NVIDIA VFX cards, Visual Enhancer executes it through the shared VFX video super-resolution path for zero-copy efficiency.
Goal Stage Stack & Key Settings Interactive Demo
Smooth High-FPS Motion (24 → 120 FPS) DLSS Frame Generation (Frame-rate mode: Target FPS, 120, Optical Flow Quality: High) Frame Generation demo
Temporal AI Video Upscaling (2× / 4×) Optional Denoising (ClearMagnetar) → Upscale (RazorNext, Upscale mode: Multiplier) Video Upscale demo
Low-Light Restoration & Grade Denoising (Noise Adaptation: On) → DLSS Neural Rendering (Style 0, Optical Flow: High) → Coloring → Sharpening (AMD CAS) Video Restoration demo
10-bit SDR-to-HDR Mastering SDR cleanup & upscaling stages above RTX Video HDR → Coloring / Grain (Animated Grain: On) + 10-bit HDR codec Color & HDR guide

Leaving every pipeline card disabled gives you a fast, GPU-optional media converter:

  • Images: Decodes source images (including camera RAW, HEIF, JPEG XL, AVIF, SVG, PNG, and TIFF) and re-encodes to your chosen format and bit depth.
  • Videos: Re-encodes or transcodes on CPU without AI/CUDA/Vulkan round-trips while preserving variable frame rate (VFR) cadence, compatible audio tracks, subtitles, and chapters.

See Export & Formats for codec, container, bit-depth, and naming options.

Aspect Preview Full Render (Render Image(s) / Render Video(s))
Scope Single loaded image or a short video segment starting from the playhead Every queued image or full video file
Video duration Configurable dropdown: 1 s, 3 s (default), 5 s, 10 s, 20 s, or 30 s Full source duration
Temporal accuracy Automatically extracts extra lookahead/history frames around the preview slice for temporal models, then trims cleanly Full continuous temporal pass
Output format Interactive viewport buffer (images) or lossless FFV1 Lossless RGB 10-bit MKV clip (videos) Your Export Settings codec, container, quality, and output directory
Realtime updates Enabled automatically when Realtime Preview is on in Settings (Coloring-only SDR images up to 8.5 MP use a fast interactive preview path) Triggered explicitly via the render button or CLI