Audio Lab — Complete Guide
This page collects the guides for 5 tools in one place. Pick a tool name from the list and its guide opens beside it, so there is no page to navigate to and no account to create.
Audio Noise Reducer
Reduce hiss, fan noise, and room tone in a recording, with a neural mode for speech and a spectral mode for everything else.
How to use it
- Choose your file. It stays on your device, and nothing is uploaded.
- Pick the mode that matches the recording: speech for voices, spectral for steady noise, music for music.
- Move the intensity slider. Listen with the A/B buttons, and use blind mode if you want an honest answer.
- Download the result as WAV, MP3, M4A, or FLAC.
Use cases
- Cleaning a podcast recorded next to a fan or an air conditioner.
- Making a voiceover usable after the room turned out to be echoing.
- Toning down hiss in an old recording before you publish it.
Limitation — Works on the recording as it is. It cannot restore a voice that was already destroyed by clipping, and the neural mode handles speech rather than music, where it produces pumping instead of a cleaner result. Long files are processed whole, which on a phone can take minutes.
Loudness Normalizer
Normalize loudness to a target LUFS, cap the true peak, and trim silence, measured with EBU R128.
How to use it
- Choose your file, then pick a preset: podcast, voiceover, video, or music.
- Build a preview first. The loudness you hear there is the loudness you will get.
- Press process and download for the real file. The tool measures the file first, then applies the gain that reaches the target.
- For several recordings, use the batch section. They come back in one ZIP.
Use cases
- Giving several episodes of a podcast the same loudness.
- Raising a quiet interview to a level people will actually finish.
- Preparing a voiceover to sit correctly next to music.
Limitation — Targets are starting points chosen from common practice, not platform rules, and the presets do not read a streaming service requirement for you. Loudness is measured over at most the first ten minutes of a file, and the true peak limit applies to the rebuilt signal rather than to the raw samples.
Audio Repair
Find clipped samples, mains hum, and clicks, then repair the ones you choose and mark them on the waveform.
How to use it
- Choose your file and press analyze. This only reads.
- Read the report: how many samples are clipped, where the hum sits, how long the silences are.
- Tick only the repairs you want, then build a preview to hear the difference.
- Process and download when the result sounds right to you.
Use cases
- Finding the exact moment a microphone was bumped during a take.
- Removing electrical hum from a recording made on mains power.
- Checking whether an old recording is damaged before spending time on it.
Limitation — Clipped material cannot be restored, only smoothed, because the information above the digital limit was never recorded. Hum repair uses the 50 or 60 Hz frequency you choose, and it does not analyse the spectrum to find the right one. Repairs are estimates, and a file with sharp real transients can come out softer than it started.
Audio Analyzer
Measure sample rate, peak, RMS, loudness, true peak, noise floor, DC offset, clipped samples, and long silences.
How to use it
- Choose your file and press run analysis.
- Read the file section for the format, and the levels section for the numbers behind the sound.
- Compare the loudness and true peak with what your platform or your player expects.
- Use the report before you change anything. It is easier to fix a problem you can name.
Use cases
- Checking the real loudness of a finished episode before you upload it.
- Finding out whether a recording is genuinely quiet or just sounds quiet.
- Verifying a file that arrived in the wrong sample rate.
Limitation — Measurements cover at most the first ten minutes of a file, and the clipped sample positions come from the first two minutes, because reading a long file at full sample rate would use a lot of memory. There is no perceived quality score, and there is no transcription: the tool measures levels, it does not judge the content.
Vocal Remover
Make a backing track by cancelling the centre channel of a stereo recording, and hear the result before you keep it.
How to use it
- Choose a stereo file. A mono recording has no centre to cancel.
- Try the "keep vocal" mode first. If that sounds clean, the voice is centred and the removal will work.
- Build a preview and compare the two versions with the A/B buttons.
- Download the instrumental, or the isolated voice if that is what you needed.
Use cases
- Making a karaoke backing track from a stereo recording.
- Separating a spoken intro from the music bed underneath it.
- Getting a dialogue-only track for editing.
Limitation — This is centre channel cancellation, not machine learning separation, so it produces one instrumental track rather than four stems, and it needs a stereo source. A voice spread across the stereo field, or drowned in reverb, leaves a trace, and centred bass and kick disappear along with the vocal.
All Loonix tools process your input in the browser: text you type and files you pick never leave your device. Anything downloaded from the internet — say the FFmpeg runtime — is a supporting binary, not your data.
