Steady fan-like noise
Recorded speech with added synthetic low hum and hiss.
48 kHz 16-bit mono WAV · 3.4 seconds
Original
Processed
Listen, compare, reproduce.
Four controlled tests using the same recorded voice. Added noise is synthetic; these are not four independent real-world recordings.
Use the same listening volume. The processed files use our current RNNoise engine with a 50% original/processed blend, with no extra mastering. The existing homepage recording remains a separate example.
Air-conditioner-like noise
The same voice with reduced volume and synthetic steady noise; not a real interview.
48 kHz 16-bit mono WAV · 3.4 seconds
Original
Processed
Light background hiss
A lightly mixed sample to check speech preservation.
48 kHz 16-bit mono WAV · 3.4 seconds
Original
Processed
Limited result: clicks
Added impulses overlap speech. This tool is not a click or clipping repair tool.
48 kHz 16-bit mono WAV · 3.4 seconds
Original
Processed
Try the same input yourself
Download an original above, open the audio tool, choose the WAV, and create a preview. These clips are shorter than 30 seconds, so the result is already the complete recording. Compare it with the provided processed file. Listen for consonants and breaths, not just a quieter background. These examples do not establish results for every voice or noise type.
Source and settings
Speech: VOiCES corpus by SRI International and Lab41, In-Q-Tel, sample sp0307/ch127535/sg0042, shared under CC BY 4.0. Dataset and attribution. We resampled the speech, adjusted its level and added deterministic noise. Download settings and file hashes.