An engine-agnostic DXGI proxy layer that intercepts frame presentation and provides a safe, extensible foundation for frame generation and temporal rendering research.
This project focuses on the hard, often-overlooked part of modern frame generation systems:
reliably entering the rendering pipeline without breaking stability.
- A DXGI / D3D12 proxy that hooks into the swapchain and presentation flow
- A robust core architecture for frame interception
- A FG-ready pipeline designed for experimentation and research
- A clean separation between infrastructure and algorithms
This repository provides the plumbing, not the algorithm.
- ❌ Not a complete Frame Generation implementation
- ❌ Not a DLSS / FSR replacement
- ❌ Not a plug-and-play FPS booster
There is no frame generation algorithm included by design.
Most experimental frame generation projects fail not because of bad math, but because of:
- fragile DXGI hooks
- broken device-lost handling
- unsafe Present / ResizeBuffers interception
- race conditions and lifecycle bugs
This project solves those problems first.
It is intended as a research base that others can build on.
- DXGI swapchain interception
- Safe Present and ResizeBuffers handling
- Device lost / reset resilience
- Thread-safe global state management
- Bypass and enable/disable control paths
- Clean API surface for future FG / temporal algorithms
- Minimal and safe
DllMaindesign
- Frame Generation research
- Temporal rendering experiments
- Optical flow–based approaches
- Engine-agnostic graphics R&D
- Learning how modern FG systems are integrated at a low level
-
Core
Lifecycle, state machine, safety guarantees -
DXGI Proxy
Swapchain interception and frame timing access -
FrameGen Interface
Algorithm-agnostic entry point (no implementation included) -
Settings & Logging
Thread-safe configuration and diagnostics
The project is structured so that algorithms can be added without modifying the proxy core.
Because algorithms are:
- subjective
- rapidly evolving
- highly experimental
The infrastructure is the hard, reusable part.
If you want to add:
- motion-vector reprojection
- optical flow
- hybrid temporal approaches
- ML-based methods
This project is meant to be the place where you start.
This project is provided as-is and may require platform-specific adjustments depending on your environment.
It is intended for developers familiar with DXGI / D3D12 internals.
MIT License.
Use it, modify it, build on it.
This project is for research and educational purposes.
It is not affiliated with NVIDIA, AMD, Intel, or Microsoft.
If you are looking for a ready-made FPS boost, this is not it.
If you are interested in how frame generation systems are wired into real rendering pipelines,
you are in the right place.