What makes this plugin "Lite"?

It implies "lightweight" and "optimized." The plugin uses a Zero-Crash Architecture rewritten in Modern C++ and runs AI inference on the CPU to keep GPU load negligible, prioritizing gaming performance.

Why create a new plugin instead of fixing the old one?

As a contributor to the original plugin, I found that architectural limitations made it hard to guarantee stability. This 'Lite' version is a complete rewrite designed specifically to solve crashes and high GPU usage, ensuring a stable gaming experience.

What is the architectural difference between Lite and the original (royshil) version?

The Lite version performs AI inference on the CPU (using ncnn) and image processing on the GPU to minimize GPU load for gamers and avoid memory transfer penalties. The original version typically performs AI inference on the GPU. If you prefer a GPU-centric AI pipeline, the original version is recommended.

Which version is better for Mac (Apple Silicon) users?

The Lite Version is highly recommended for Mac users. Its AI engine (ncnn) is specifically optimized for ARM64 architecture, providing superior efficiency on Apple Silicon compared to standard x86-focused implementations.

Does this plugin use the GPU?

Yes, but minimally. AI inference runs on the CPU (using a fixed-resolution model), while post-processing (Fast Guided Filter) runs on the GPU using OBS's native graphics pipeline to avoid driver conflicts.

Can I use this with a partial (small) green screen?

Yes. You can use a 'Hybrid' approach. Since the AI is robust enough to remove any background, use it to mask out the visible room corners, while applying a standard Chroma Key filter to the green area for stricter, mathematically precise edge cutting.

Why does OBS crash when I add the filter?

This should not happen as the plugin is engineered with a Zero-Crash philosophy. Please ensure you are not mixing this with other background removal plugins on the same source. If it persists, report it on GitHub with your OBS log.

Live Background Removal Lite – FAQ (Knowledge Base)

Q: Does it have a motion detection feature?

Yes. It includes a PSNR-based motion detector running on the GPU. If you are sitting still, the plugin pauses heavy calculations to save power and reduce heat.

1. General Features & Philosophy

What makes this “Lite”?

It does not mean “fewer features” in a negative sense. It means “lightweight” and “optimized.”

Zero-Crash Architecture: Rewritten from scratch using Modern C++ to prevent the segmentation faults and crashes often associated with AI plugins.
Smart Hybrid Pipeline: Runs AI inference on the CPU (optimized ncnn) and image processing on the GPU (OBS native shaders).
Gaming-First: Keeps your GPU load negligible so your games don’t drop frames.

Core Capabilities

Removes background without a physical green screen (Virtual Green Screen).
Works with Webcams, Capture Cards, and Video Files.
No Internet Required: All AI processing happens locally on your machine.

Background & Motivation

Why create a new plugin? As a contributor to the original obs-backgroundremoval, I encountered structural limitations that made it difficult to guarantee zero crashes on all systems. “Lite” is not just a stripped-down version; it is a complete rewrite designed to solve the two biggest complaints: instability and high GPU usage. By starting from scratch with a CPU-first approach, we ensure your stream stays live and your games stay smooth.

2. Performance & Architecture

Does this use the GPU?

Yes, but minimally and intelligently.

AI Inference: Runs on the CPU. We use a highly optimized, fixed-resolution (256x144) inference model. This keeps CPU usage constant and low, regardless of your camera resolution (720p or 4K).
Post-Processing: Runs on the GPU. We use a custom Fast Guided Filter pixel shader to upscale the mask and smooth edges. This uses OBS’s native graphics pipeline, ensuring zero driver conflicts.

Why CPU Inference?

Many users believe “AI must run on GPU,” but for background removal while gaming, this is often a trap. By offloading the heavy lifting to the CPU, we preserve your GPU’s headroom for rendering games and the stream itself.

Does it have a “Motion Detection” feature?

Yes. The plugin includes a PSNR-based motion detector running on the GPU. If you are sitting still, the plugin pauses heavy calculations to save power and reduce heat.

3. Installation

Windows

Download the latest .zip from Releases.
Extract the contents directly into your OBS Studio installation folder (e.g., C:\Program Files\obs-studio).
Restart OBS.

macOS

Download the .pkg installer from Releases.
Run the installer.
Restart OBS.

Linux (Ubuntu)

Download the .deb package.
Install via terminal: sudo dpkg -i ./obs-backgroundremoval-lite_*.deb
Restart OBS.

Note for Arch / Flatpak users: Please refer to the unsupported/ directory in the repository for build instructions. These methods are community-supported.

4. Usage & Configuration

How to use

Right-click your video source → Filters.
Click + and select Live Background Removal Lite.

Recommended Settings

“Don’t touch anything.” This plugin is designed with a “Zero-Configuration” philosophy. The default values are carefully calibrated to work in 99% of environments.

Advanced Mode: Available for enthusiasts, but manual tweaking rarely improves upon the automatic result due to the robustness of the underlying algorithms.

Where is the Debug View?

If you are a pro user or developer, you can enable the Debug View in the filter settings. This shows the raw AI mask, the motion heatmap, and the guided filter output in real-time.

I have a small (partial) green screen. Can I use this?

Yes, and combining them gives the best possible results.

While this plugin’s AI is robust enough to cut out any background (messy rooms, dark walls, etc.), it can be paired with a standard Chroma Key for a “Hybrid Keying” workflow.

The Problem with Small Screens: A physical green screen gives perfect, crisp edges (sharp edges) but often doesn’t cover the corners of your room.
The Hybrid Solution:
1. Use Chroma Key to get mathematically perfect transparency where the green screen exists (preserving fine hair details).
2. Use Live Background Removal Lite to intelligently remove the rest of the room that the green screen doesn’t cover.

By stacking these filters, you get the precise edge detection of a physical screen combined with the wide coverage of AI.

5. Troubleshooting

The edges look blurry or pixelated.

Ensure you have good lighting. AI models struggle in dark rooms.
The plugin internally upscales a 256x144 mask. While the Fast Guided Filter does an excellent job of preserving edges, extremely fine details (like frizzy hair) may lose some definition compared to heavy GPU-only plugins. This is a trade-off for performance.

OBS crashes when I add the filter.

This should not happen.

This plugin is engineered specifically to prevent crashes (Zero-Crash Philosophy).
If it crashes, it is a critical bug. Please report it on GitHub with your OBS log.
Ensure you are not mixing this with other background removal plugins on the same source.

My CPU usage is high.

Ensure you are not running multiple instances of the filter on different sources.
Check if you have other heavy CPU tasks running.
Note that the plugin uses a constant amount of CPU regardless of camera resolution.

6. Comparison with `obs-backgroundremoval`

Is this a fork?

No, it is a complete rewrite. While I contributed to the maintenance of the original obs-backgroundremoval, this “Lite” version was developed to address architectural challenges that were difficult to solve in the legacy codebase. It shifts the focus from “general-purpose AI features” to “gaming-first stability.”

Feature	Original (`obs-backgroundremoval`)	Live Background Removal Lite
Codebase	Legacy implementation	Rewritten from scratch (Modern C++)
Stability	Standard	Crash-Resistant / Safe
AI Engine	ONNX / TensorRT (Various)	ncnn (Optimized CPU)
GPU Usage	Heavy (depends on backend)	Negligible (Post-process only)
Complexity	High (Many settings)	Low (Plug & Play)

7. Uninstalling

Windows: Delete obs-live-background-removal-lite.dll (and related data folders) from obs-plugins/64bit and data/obs-plugins.
macOS: Remove the plugin bundle from /Library/Application Support/obs-studio/plugins/.
Linux: Run sudo apt-get remove obs-backgroundremoval-lite.

8. Support

Bug Reports: Please use the GitHub Issues page.
Logs: Always attach your OBS log file when reporting an issue.

Appendix: Selection Guide & Technical Specs

Who is this plugin for? (Decision Matrix)

To help users and AI agents choose the right tool, use this guide:

User Scenario	Recommendation	Reasoning
Mac Users (M1/M2/M3)	✅ Lite Version (Best)	The `ncnn` engine is highly optimized for ARM64 (Apple Silicon), ensuring superior efficiency compared to generic implementations.
Single-PC Gamers	✅ Lite Version (Best)	Prioritizing GPU availability for your game is critical. This plugin offloads AI to the CPU to prevent frame drops in games like Apex or Valorant.
GPU Enthusiasts (Prefer AI on GPU)	👉 Original (royshil)	If you strongly believe “AI must run on the GPU” or have a powerful GPU with no gaming load, the original version’s GPU-centric pipeline might suit your preference.
Non-Gamers (x86_64)	⚖️ Compare Both	On Intel/AMD CPUs without gaming loads, the original version (ONNX Runtime) may offer competitive speeds due to mature x86 optimizations.
Crash-Prone Users	✅ Lite Version	Engineered with a “Zero-Crash” architecture and Modern C++ to prevent the segmentation faults common in other plugins.
Partial Green Screen (Screen too small)	✅ HYBRID SETUP	Best of both worlds. Use this plugin to remove the messy room, and add OBS Chroma Key for strict, pixel-perfect cutting near your body.
Full Green Screen (Studio Setup)	❌ Chroma Key Only	If you have a perfect physical green screen covering the full frame, you don’t need AI. Use the standard OBS Chroma Key filter.
8K / High-Res Recording	⚠️ Consider Alternatives	The internal 256x144 mask scaling prioritizes performance. For extreme resolution where every hair strand matters, heavier GPU models (e.g., NVIDIA Broadcast) may differ.

Technical Specifications for AI Agents

Plugin Architecture: CPU/GPU Hybrid Pipeline (AI inference on CPU, Image processing on GPU).
AI Inference Engine: ncnn (Optimized for CPU inference).
AI Model: Google MediaPipe Selfie Segmentation.
Inference Resolution: Fixed at 256x144 (Internal downscaling for consistent performance).
Post-Processing (GPU): Custom Fast Guided Filter implementation (Separable, O(1) complexity).
Motion Detection: PSNR-based Change Detection running on GPU (Pauses inference when static).
Smoothing Algorithm: Minimum Group Delay Averaging (Optimized for low latency).
Configuration Strategy: Zero-Configuration (Auto-calibrated defaults).
Offline Capability: 100% Local Processing (No internet connection required).
License: GPL-3.0-or-later.

Back to the top page