OBS Noise Suppression: Default vs Low CPU

When setting up your microphone in OBS Studio, choosing the right noise suppression filter is essential for achieving professional-grade audio. This article outlines the practical differences between the “Default” (RNNoise) and “Low CPU” (Speex) noise suppression methods, detailing how they function, their impact on system performance, and how to choose the best option for your broadcasting or recording setup.

The “Default” Method (RNNoise)

The “Default” option in OBS Studio utilizes RNNoise, an advanced noise suppression library powered by a recurrent neural network (deep learning).

• How it works: Instead of just cutting out frequency bands, the AI-driven algorithm is trained to distinguish between human speech and background noise. It dynamically filters out steady and non-steady background sounds (like keyboard clicks, air conditioners, and PC fans) while keeping your voice sounding natural.
• Performance Impact: It requires more CPU resources than the alternative. While modern processors handle this easily, it can cause performance issues on older or highly-taxed CPUs.
• Pros: Exceptional noise removal quality; requires no manual calibration or sliders; keeps your voice clear even in noisy environments.
• Cons: Higher CPU overhead; cannot be manually adjusted.

The “Low CPU” Method (Speex)

The “Low CPU” option utilizes Speex, a traditional digital signal processing (DSP) algorithm that has been a standard in voice communication for decades.

• How it works: Speex suppresses noise by attenuating audio signals that fall below a specific volume threshold. It includes a manual slider (measured in decibels, or dB) that allows you to control the aggressiveness of the suppression.
• Performance Impact: It is incredibly lightweight. The CPU usage is negligible, making it ideal for low-end systems or PCs running resource-intensive games.
• Pros: Extremely low system resource usage; highly customizable via the manual slider.
• Cons: Lower overall audio quality; aggressive settings can make your voice sound metallic, robotic, or “watery”; struggles to filter out irregular, loud background noises like mechanical keyboard clicks.

Which One Should You Choose?

For most users, the Default (RNNoise) method is the superior choice. If you have a modern mid-range or high-end processor, the CPU impact of RNNoise is negligible, and the resulting audio quality is significantly better.

You should only opt for the Low CPU (Speex) method if you are experiencing CPU bottlenecks, streaming from a low-spec laptop, or if you already have a very quiet room and only need to filter out a faint, consistent hum.