OBS Audio Settings: Simple Clean Voice Without Sounding Weird
Bad audio kills streams faster than bad video. People will tolerate a slightly soft image, but they leave when your voice is muffled or robotic.
This guide keeps it simple.
Table of Contents
- The goal for voice audio
- Common settings that ruin sound
- A simple clean setup
- Loudness and clipping
- 24/7 stream notes
- Recommended filter chain
- Monitoring and quick fixes
1. The goal for voice audio
Clear, consistent volume, no harsh noise suppression.
2. Common settings that ruin sound
Over-aggressive noise suppression can make voice robotic.
3. A simple clean setup
Use modest filters and test recordings.
4. Loudness and clipping
Clipping sounds terrible. Keep headroom.
5. 24/7 stream notes
Stability matters. If you stream 24/7, consider a server.
See /streaming.
6. Recommended filter chain
For clean voice with minimal artifacts:
- Noise Suppression: Light
- Compressor: Low ratio
- EQ: Gentle shaping
- Limiter: Prevent clipping
Test with short recordings and adjust slowly.
7. Monitoring and quick fixes
Monitor for distortion, pumping, or robotic tone. If voice sounds strange, reduce suppression strength and lower compressor ratio.
Audio chain that works for spoken-word streams
For a 24/7 talk / news / lofi-with-DJ stream, three things separate "fine" from "broadcast clean":
- Sample rate / bit depth consistent throughout.
- Noise suppression BEFORE compression.
- Loudness target matched to the platform (-14 LUFS for YouTube, -16 LUFS for Twitch).
OBS settings that matter (Settings > Audio)
| Setting | Recommended | Why |
|---|---|---|
| Sample Rate | 48 kHz | matches video frame timing |
| Channels | Stereo | platform standard; mono ok if voice-only |
| Mic monitoring | off (during stream) | feedback risk |
| Audio bitrate (output) | 160 kbps AAC | YouTube cap; higher is wasted |
Twitch caps audio at 160 kbps even if you set 320. YouTube allows 384 but no listener equipment will reveal it.
Filter chain on the mic source
Right-click mic source > Filters. Order matters:
- Noise Suppression (RNNoise) - removes hiss, fan noise.
- Noise Gate - threshold ~ -45 dB, attack 5 ms, release 150 ms. Cuts silence between sentences.
- Compressor - ratio 4:1, threshold -24 dB, attack 6 ms, release 60 ms, output gain +6 dB. Levels peaks.
- Limiter - threshold -1.5 dB, release 60 ms. Prevents clipping.
- Gain (optional) - last in chain to hit your target.
Loudness target by platform
| Platform | Target | Peak ceiling |
|---|---|---|
| YouTube | -14 LUFS | -1 dBTP |
| Twitch | -16 LUFS | -1 dBTP |
| Spotify (rebroadcast) | -14 LUFS | -1 dBTP |
| Internal monitoring | -23 LUFS (EBU R128) | -1 dBTP |
Use Youlean Loudness Meter (free OBS-compatible) to measure. Run a 1-minute test, adjust gain stage so live LUFS hovers at target.
Common setup mistakes
| Mistake | Effect | Fix |
|---|---|---|
| Compressor before noise suppression | hiss gets boosted | put suppression first |
| No limiter | one loud word pegs at 0 dB and clips | always last filter |
| Mic monitoring on during stream | echo, then feedback loop | turn off during live |
| Sample rate mismatch (44.1 vs 48 kHz) | drift, occasional click | force 48 kHz everywhere |
| Mic plugged into front-panel jack | electrical noise from PSU | use rear or USB interface |
Hardware that punches above its price
For talk streams, you don't need a $500 condenser. The proven cheap stack:
- Samson Q2U / Shure MV7+ (USB) - dynamic, rejects room noise.
- Boom arm + shock mount - lifts mic close to mouth, stops desk vibrations.
- Foam pop filter - kills plosives.
- Closed-back monitor headphones - any will do; no monitoring through speakers.
Listener perspective check
After setup, listen back through different devices:
- Phone speaker (worst case).
- Cheap earbuds.
- Decent headphones.
If voice is intelligible on all three at 50 % volume, your loudness and EQ are right. If you have to crank the phone, your average loudness is too low; if it distorts on headphones, your peaks are clipping.
