
Try the OBS Expander's RMS detector when brief keyboard clicks open the microphone too readily, while keeping the rest of the setup fixed for comparison. RMS averages a short interval; Peak reacts more directly to momentary level changes. Neither detector is a guarantee that typing will disappear while you speak.
Key Takeaways
- Detector mode decides how the input level is measured.
- Threshold and ratio decide how much quiet material is reduced.
- Preserve quiet words while testing unwanted short peaks.
Understand which decision the detector makes
The Expander reference describes RMS as an average over the preceding 10 milliseconds, which reduces sensitivity to quick peaks, and Peak as a faster response to level changes. It also distinguishes an expander's adjustable ratio from a gate-like response.
Record the current mode, threshold and ratio before changing anything. A keyboard click and a quiet consonant can both be short, so raising the threshold until the meter stops moving can also remove wanted speech. Detector choice is a separate variable worth evaluating first.
Compare the same performance
Make two short test recordings of a consistent sequence: quiet speech, normal speech, typing during a pause and typing during speech. Change only the detector mode between them. Keep microphone position and input gain stable so you are comparing the intended control.
Listen to complete words rather than judging only the silence between sentences. RMS may avoid opening on some brief peaks, but speech that already opens the expander also allows other simultaneous sounds through. If clicks remain during talking, do not claim that a higher ratio can identify keyboard sounds independently of the voice.
Proposed caption-recording session
Suppose a creator records spoken explanations while entering short commands. The desired result is natural narration with less distracting idle keyboard noise. Compare the command names, sentence starts and final syllables in both recordings, then compare typing in the gaps.
Choose the mode that best fits that evidence. If both modes damage the narrator, revisit the microphone arrangement or reduce the amount of expansion instead of stacking more aggressive settings. The official guidance warns that an excessive threshold can cut off the voice.
Make the acceptance rule audible
Accept the configuration only when the actual spoken terms remain intelligible and the targeted background behavior improves enough for the use case. Document any remaining typing under speech rather than describing the result as isolated voice. This decision concerns level detection, not restoration of clicks already baked into a recording. Keep the unprocessed source or original recording available if later editing needs a different approach.
Related tasks
- Why Does a Noise Gate Cut Off Quiet Words and Sentence Endings?
- How Do I Remove Mouth Clicks Without Damaging T and K Sounds?
- Why Does Background Noise Swell Between Words After Voice Compression?
Sources and scope
- Expander Filter: RMS averages the preceding 10ms to reduce sensitivity to short peaks; Peak responds faster; threshold and ratio affect below-threshold reduction.
OBS project documentation read directly through the web tool on 2026-10-23 supplies technical support. Community posts, where present, show individual demand only. Examples and listening checks are proposed procedures; no actual recording or repair is claimed.
Prepared by Imagild Editorial, the company editorial team responsible for these guides. AI assisted research and drafting. This is documentation-based guidance with proposed examples, not a claim of hands-on testing or an independent product review. Report a correction.
Cover: Rumeyda Tutak / Pexels, Pexels License. This is an illustrative source photograph, not an application screenshot or a generated test result.