Share your music, connect with artists, and explore new content. Upload your sounds, music, and podcasts to the world.
Clean audio gives listeners more of what they came for: the DJ’s track choices, the producer’s arrangement or the podcast’s conversation. Hiss from an air conditioner, mains hum, traffic rumble, keyboard clicks and room reflections can all make a recording harder to enjoy, particularly when it is played through earphones.
Work from the original recording, make a duplicate and apply changes in small, reversible steps. Background noise is only one part of audio restoration; echo, clipped speech, harsh sibilance and uneven volume need their own treatments. Treating those problems separately preserves a natural sound instead of producing the thin, over-processed result that aggressive denoising often creates.
Start by describing the unwanted sound. Steady hiss usually comes from airflow or an electronic noise floor; a low buzz often follows the 50 Hz mains frequency used in Australia; and distant traffic creates slow, uneven rumble. Clicks and crackles may be transient defects, while a neighbour’s voice is interference that may need editing rather than denoising.
Listen on headphones and ordinary speakers, then record 30 to 60 seconds of room tone with the microphone, gain and monitoring setup unchanged. For repeatable comparisons, use the same short reference recording before and after processing, and choose profile samples containing the noise but no speech, music or traffic swell. A section with a passing car may train the software to suppress far more than intended.
Many clean recordings begin with ordinary precautions. Keep the microphone close to the speaker, use a pop filter for plosives and record in a soft, furnished room. A home setup in Perth may pick up refrigerator hum, a Melbourne flat near a tram route may add low-frequency rumble, and a Brisbane studio on a hot day may need the air conditioner running. Each needs a different solution.
Australian conversational speech can contain sharp “t” sounds, bright “s” consonants and abrupt breaths. Leave a little space before and after each take, watch the level meter and avoid speaking over the microphone’s handling noise. Capture 24-bit WAV at 48 kHz if your equipment supports it, back up the raw file, and resist noise reduction while recording; a clean source gives the editor much more room to work.
Noise reduction works best when the unwanted sound is predictable. In Audacity, Adobe Audition or iZotope RX, use the relevant noise-profile or capture function to select a few seconds of steady background sound. A clean section of room tone is ideal. Include the hum if it is steady, but exclude clips containing voices, cymbals or a passing vehicle because those features would be learned as noise.
Preview the result continuously and begin with a conservative reduction, roughly 6 to 12 dB, rather than the maximum slider setting. Move through the entire recording at normal speed and in short sections. Listen for musical “underwater” artefacts, metallic edges on consonants, unnatural gaps between words and noise that rises and falls when the software mistakenly detects speech.
Some files need a different pass. Apply de-hum filtering to electrical buzz and a gentle high-pass filter to deep rumble, using the narrowest useful frequency range. Repair individual clicks or pops with spectral tools or a short crossfade, and tame sibilance with de-essing rather than lowering the whole vocal level.
For dialogue, edit separate words and breaths instead of denoising the entire programme as one block. Crossfade around cuts, retain a little room tone so the voice does not sound pasted onto the background, and check that lip sounds and mouth clicks have not been removed with the noise. When two remote contributors were recorded separately, align and level their tracks before joining them.
For a DJ mix, leave crowd noise, vinyl crackle and ambient sound if they belong to the performance. A continuous reduction can weaken bass, dull cymbals and flatten the movement that makes a live set feel alive. If stems or source tracks are available, clean them individually, rebuild the balance with light processing and leave the final blend mostly untouched. That avoids reprinting the original hiss.
Fine-tune with equalisation and light compression after the main noise reduction, not before. Too much compression can turn a steady hiss into audible pumping, while broad high-frequency cuts can make speech sound muffled. Set levels for clarity and musical impact, then make a separate pass for loudness. A clear explanation of loudness normalization shows how consistent delivery levels differ from simply turning up a quiet recording.
For spoken-word material, a loudness target around -16 LUFS can be a useful starting point, but the right setting depends on the platform, genre and dynamics. Leave suitable headroom and check true peaks rather than relying on the peak meter alone. For music, preserve the intended contrast between quiet introductions and full sections. The goal is consistent playback, not every bar having the same apparent volume.
Check the finished file on studio monitors, earphones, earbuds and a phone speaker. Bluetooth speakers often conceal subtle artefacts that appear on a club system, while earphones reveal clicks and sibilance that a phone may hide. Pause in several places and listen for a low hum, a sharp edge on consonants or an unnatural amount of silence around breaths.
Keep the untouched recording and export the cleaned version as a lossless WAV, commonly 24-bit at 48 kHz. Do not create an MP3 and upload that as the only master; lossy compression can colour cymbals, vocal sibilance and quiet noise that was painstakingly removed. Name the files clearly, including the processing version, so they remain identifiable in your archive.
Compare the original and processed files at matched volume, then listen to the complete export without skipping ahead. Look for noise pumping, metallic resonance, clipped peaks, abrupt room-tone changes and missing word endings. A spectrogram can expose persistent hiss or harsh cut edges, but the ears should decide whether the restoration sounds convincing.
Mix-Sets hosts the final audio, so the noise-removal work should happen in your chosen editor before upload. The Pro plans offer more upload space and advanced features for a growing catalogue. Keep masters separate from previews, preserve the raw take locally, and check the uploaded file on another device before sharing it.
Once the file is ready, describe it accurately and organise it so the right listeners can find it. A useful title, genre tags and playlist placement matter more than claiming a set is perfectly clean. You can also publish a short before-and-after excerpt to demonstrate the restoration, provided the original recording and any voices involved are appropriate to share.
Browsing fresh audio on Mix-Sets can help you compare presentation, tagging and track selection with other creators, while audience statistics show where people stop listening. Australian listeners may encounter your work through Triple J, community radio, dance floors, social clips or everyday earbuds, so test the master in several of those contexts. Preserve enough dynamics for a club system without making speech feel remote on a phone.
The most reliable method combines a careful recording, a representative noise profile and restrained editing, with frequent checks on different playback systems. Save every original so a later decision can be reversed. What should remain in the listener’s memory is the performance, not the machinery used to clean it.
