OBS Studio RNNoise vs Speex Noise Suppression
When setting up audio in OBS Studio, choosing the right noise suppression filter is crucial for achieving clear, professional sound. This article compares the two primary built-in methods, RNNoise and Speex, highlighting their differences in technology, CPU performance, audio quality, and configuration options to help you decide which filter is best for your streaming or recording setup.
Speex (Low CPU Usage, Algorithmic)
Speex is a traditional, frequency-based noise suppression method that has been a staple in audio processing for many years.
- How it works: It uses standard digital signal processing (DSP) algorithms to detect and suppress constant, predictable background frequencies, such as the steady hum of a computer fan or air conditioner.
- CPU Impact: Extremely low. Speex requires almost negligible processing power, making it the ideal choice for budget setups, older hardware, or resource-heavy gaming streams where every frame counts.
- Control: It features a manual slider allowing you to adjust the suppression level in decibels (dB). Users can fine-tune exactly how aggressive the filter is.
- Drawbacks: Speex is not effective against sudden, non-stationary sounds like keyboard clicks, mouse taps, or barking dogs. Furthermore, turning the suppression level too high often results in a “tinny” or robotic voice.
RNNoise (High Quality, AI-Powered)
RNNoise is a modern noise suppression method based on deep learning and artificial intelligence.
- How it works: It utilizes a recurrent neural network (RNN) trained on thousands of hours of audio to distinguish human speech from background noise. It automatically detects and filters out unwanted sounds in real-time.
- CPU Impact: Moderate. Because it relies on AI algorithms, RNNoise uses more CPU resources than Speex. While negligible on modern multi-core processors, it may cause performance drops or audio stuttering on older, low-end PCs.
- Control: It is completely automated. There are no sliders or configuration options; you simply turn it on, and the neural network does the work.
- Audio Quality: Exceptional. RNNoise is highly effective at removing complex, unpredictable noises like keyboard typing, paper rustling, and traffic, while keeping your voice sounding natural and undistorted.
Key Differences At a Glance
| Feature | Speex | RNNoise |
|---|---|---|
| Technology | Classical DSP Algorithm | AI / Neural Network |
| CPU Usage | Extremely Low | Moderate |
| Voice Quality | Average (can sound tinny) | High (natural voice) |
| Dynamic Noise | Poor (struggles with clicks) | Excellent (filters clicks, fans, etc.) |
| Configuration | Manual slider control | Fully automated |
Which One Should You Choose?
Choose RNNoise if: * You have a modern computer with spare CPU capacity. * You have an environment with unpredictable noises like mechanical keyboards, traffic, or family members talking in the background. * You want a “set-and-forget” solution with the highest possible audio quality.
Choose Speex if: * You are streaming or recording on a low-end or older PC, and OBS is struggling with CPU usage. * Your background noise is a constant, steady hum (like a fan) that can easily be filtered out with manual tuning. * You want precise, manual control over the exact decibel level of your noise suppression.