Windows Desktop App

PixelForge AI

A cinematic, local-first video upscaling workflow for creators who need sharper exports, cleaner restores, and faster quality checks before committing to a long render.

Windows 10/11 Vulkan GPU required 3 upscaling engines + RIFE

PixelForge AI Video Enhancer

Preview first. Commit later.

v1.0.16
Before
After
Engines Real-ESRGAN, RealSR, Waifu2x
Interpolation RIFE GPU + minterpolate CPU fallback
Estimate Stage ETA + hardware-aware guidance
Local

Your footage stays on your own machine instead of being uploaded to a hosted render service.

Compare

Draggable before/after preview helps you check results before committing to a full pass.

ETA

Stage-by-stage processing estimates highlight the slowest parts of the pipeline up front.

Update

Built-in GitHub release updates reduce friction after the initial install.

Built For Real Video Cleanup

Inspired by premium video-tool landing pages, but positioned around this app's actual desktop workflow and feature set.

Three AI Upscaling Engines

Real-ESRGAN for general footage, RealSR (GPU) for photo-realistic DF2K upscaling, and Waifu2x (GPU) for anime and illustration content — all selectable from one interface.

RIFE GPU Frame Interpolation

Real-time frame interpolation via ncnn-vulkan using RIFE v4.6 (recommended) or RIFE v4 (legacy). CPU fallback via minterpolate available for non-Vulkan setups.

Runtime Awareness

ETA guidance uses hardware detection and stage timing profiles so you can test with shorter clips before long exports.

Direct Auto-Updates

The updater checks GitHub Releases, downloads the latest Windows build, and reduces manual maintenance after setup.

Workflow Profiles

Fast drafts when you need answers. Higher detail when you are ready.

PixelForge AI is structured around practical production behavior: run a quick check first, inspect a compare view, then move into a cleaner pass once the settings are proven.

Quick Preview Faster test renders for early decision making.
Balanced Workflow A default middle ground for quality and runtime.
Max Detail Higher-fidelity output settings for final passes.

Upscaling Engines

  • Real-ESRGAN — general footage and restoration
  • RealSR (GPU) — photo-realistic upscaling with DF2K and DF2K_JPEG models
  • Waifu2x (GPU) — anime and illustration with cunet and upconv_7 models

Interpolation & Advanced Controls

  • RIFE GPU — real-time frame interpolation via ncnn-vulkan (rife-v4.6 recommended)
  • minterpolate CPU — fallback for non-Vulkan systems
  • Model selection, scale, thread tuning, denoise, sharpen
  • CRF, audio, target FPS, output dimensions, clip range

How The App Works

A straightforward desktop workflow built around reducing wasted processing time.

01

Select input and output

Load a source file, choose where the export should land, and define a short clip range when you want a quick validation pass.

02

Pick a profile

Choose a speed profile and a content profile so the app sets sane defaults before you go deeper into advanced controls.

03

Preview and estimate

Use the compare preview and ETA panel to understand how the current settings affect quality, runtime, and the expensive stages.

04

Run the final pass

Launch the full job once the settings are proven, then rely on local processing and bundled runtime assets to complete the export.

Packaging

Packaged As A Standalone Windows App

The repository ships a ZIP containing all four ncnn-vulkan engines and their model folders. A GPU with Vulkan support is required for Real-ESRGAN, RealSR, Waifu2x, and RIFE processing. The in-app updater stages updates through AppData and relaunches cleanly without requiring a manual reinstall.

System requirement

A GPU with Vulkan support is required for GPU-based upscaling and RIFE interpolation. minterpolate is available as a CPU fallback for interpolation only.

Need A Custom Media Or Automation Workflow?

Knight Logics builds local-first tools, practical desktop utilities, and automation systems around real operator workflows.