How to Remove Background Noise in OBS
By the Inverse.AI team · Updated June 2026
Quick answer: In OBS Studio, click the gear next to your mic in the Audio Mixer → Filters, click +, add Noise Suppression, and choose RNNoise (better quality) over Speex (lighter on CPU). That's it for live streams. RNNoise is great for steady noise in real time, but it's a lightweight model — for recorded video where quality matters more than latency, clean the exported recording with a stronger tool like Noise Reducer.
OBS Studio is the standard for streaming and screen recording, and it has built-in noise suppression that runs in real time. Here's how to set it up, which method to pick, and why your recordings deserve a second pass.
Step-by-step: add the Noise Suppression filter
- In the Audio Mixer, click the ⋮ (or gear) next to your microphone source → Filters.
- Click + → Noise Suppression.
- Choose a Method:
- RNNoise — a small recurrent neural network; the best built-in quality, modest CPU cost. Pick this for most setups.
- Speex — older DSP method, lighter on CPU; adjustable suppression level in dB.
- (Optional) Add Noise Gate below it to mute the gaps between speech, and Compressor to even out levels.
Recommended live chain
Noise Suppression (RNNoise) → Noise Gate → Compressor → Limiter. Order matters: suppress noise first, then gate the silence, then control dynamics.
Live vs recorded — the key trade-off
RNNoise is optimized to run in real time with low latency, which means it's deliberately lightweight. For a live stream that's the right call. But for a recorded video or podcast — where you can afford a few seconds of processing for much better quality — the live filter is leaving quality on the table.
Where OBS noise suppression hits its limits
- Heavy or non-stationary noise (construction, traffic, overlapping voices) — RNNoise smooths it but won't fully remove it.
- Wind and strong room echo — not its strength.
- Quality ceiling — being real-time-lightweight, it can't match an offline model on a finished file.
When to clean the recording online instead
For finished recordings, run the exported file through a model with no real-time constraint. Noise Reducer uses a custom Mel-band neural network with a LoRA adaptation layer to separate voice from noise far more thoroughly than a live filter can — removing wind, traffic, hum and chatter while keeping speech natural. It accepts your video recording directly (MP4, MKV, MOV) and returns it cleaned in ~30 seconds: remove noise from video. Curious about the model? See how our noise-reduction AI works.
Best practice: RNNoise live for the stream → Noise Reducer on the VOD/recording before you publish it.
Frequently asked questions
RNNoise or Speex in OBS — which is better?
RNNoise gives better quality (it's a neural model); Speex is lighter on CPU. Use RNNoise unless you're CPU-constrained.
Why is there still background noise after OBS Noise Suppression?
RNNoise is a lightweight real-time model — it can't fully remove heavy, dynamic, or windy noise. Clean the recording afterward with Noise Reducer.
Can OBS remove noise from a recording after the fact?
OBS filters apply live. To clean an existing recording, upload it to Noise Reducer.
Does Noise Suppression add latency to my stream?
RNNoise adds a small amount; for most setups it's negligible. If latency is critical, Speex is lighter.
From the makers of Noise Reducer — 5M+ downloads, 4.6★.