In this guide
A channel is audio already stored in a file. A stem is an audio part grouped by its source or role. Check the left and right channels before requesting AI separation: your wanted recording might already be on one side.
Hear what a stereo-to-mono mix changes
These original synthesized tones demonstrate channel arithmetic. They are not AI separation results. Compare the original stereo file with its left, right, and averaged mono exports.
Start at a comfortable playback volume and keep the player volume consistent. Your device may already downmix stereo; headphones make the left-only placement easier to hear. Opposite-polarity stereo can also cancel when your playback system mixes it to mono.
Example 1Matching left and right channels
Both channels contain the same tone, at the same sample amplitude and polarity. Averaging them keeps that amplitude: (L + L) / 2 = L.
What to do with your recording: When two channels really are identical, the average does not reduce their sample level. You do not need to turn it up to compensate for the channel count.
Example 2A tone on the left, silence on the right
The right channel contains zero-valued samples. Averaging halves the left signal: (L + 0) / 2 = L / 2. The mono mix has a sample peak 6.02 dB below the left-channel export.
What to do with your recording: For a wanted recording on one side and an empty other side, extracting the useful channel avoids this level reduction. That is a channel choice, not a noise-removal process.
Example 3Equal channels with opposite polarity
Every right-channel sample is the negative of its left-channel counterpart. Their average is zero: (L + −L) / 2 = 0. Both separate channels contain a tone, while this mono mix is silent.
What to do with your recording: A silent mix does not prove the original file was empty. Check the separate channels before discarding a recording. Real microphone recordings can have partial cancellation that varies by frequency and time.
Each file is 3 seconds of 48 kHz, 16-bit PCM WAV. Sources use a 440 Hz tone with a peak of 0.25 (−12.04 dBFS) and 20 ms edge fades. Mono exports use the free splitter’s channel selection and (left + right) / 2 averaging, without normalization. These are sample-level comparisons; playback loudness depends on your device and channel routing.
Download an Original stereo WAV and open it in the free channel splitter to try the choices yourself. Use the WAV inspector to check the channel count and format.
Audio channels and stems describe different things. A stereo file has left and right channels, but those channels are not automatically vocals and instruments. Understanding that difference can save an unnecessary separation attempt—or prevent you from throwing away a useful part of a recording.
Start by asking how the recording was made. Did two microphones feed separate channels? Did a screen recorder place different inputs on different sides? Or is this a finished stereo mix with most sounds spread across both? The answer determines whether channel extraction, a mono mix, or source isolation fits the task.
What is an audio channel?
A channel holds one stream of samples. Mono audio has one channel; stereo has two. “Left” and “right” describe their conventional playback positions. A person, an instrument, and room sound can all be mixed into either channel.
A two-channel recorder may store two microphones separately, but that is a recording choice rather than a promise made by the WAV or MP3 format. Likewise, a mono voice played through two speakers is still a one-channel recording. The output device does not create a second recorded source.
What is a stem, and how is it different?
In an editing workflow, a stem is a separately usable part grouped by its source or role, such as vocals, drums, or dialogue. A stem can itself be mono or stereo. Its defining feature is what it contains, not its channel count.
If the original production tracks are available, use those when they meet your goal. When only a finished mixture is available, AI source separation estimates a requested source from it. Selecting the left channel simply keeps audio that was already on the left; it does not decide which samples belong to a singer.
| Operation | What it does | What it cannot assume |
|---|---|---|
| Extract left or right | Keep one existing channel as a mono file | That channel contains only the wanted speaker or instrument |
| Mix stereo to mono | Combine both channels into one stream | The combination keeps every wanted sound at the same level |
| AI source isolation | Estimate a described sound from a mixture | The result matches an original production track perfectly |
When extracting a channel is enough
Suppose a recorder placed a presenter’s microphone on the left and a guide track on the right. If the presenter’s channel contains the performance you need, you can export that side directly. There is no reason to ask a separator to reconstruct a voice already available in its own channel.
Use the free audio channel splitter to open a short stereo MP3 or WAV. Prepare Left channel, listen, and save it if useful. Then prepare Right channel and check that side too. Each download is a mono WAV containing the selected channel’s full decoded duration.
Do not infer quality from the larger waveform alone. A louder right channel may be a noisy room microphone, while a quieter left channel contains clean close-microphone speech. Listen to quiet words, loud passages, and the difficult overlap before deciding which recording to keep.
Audacity’s stereo splitting documentation describes separate channel tracks and center-panned mono tracks. In a desktop editor, you can retain both for independent editing. With Saturalabs’ free splitter, download each side separately if you need both files.
What to do with a voice on only one side
If your voice appears only on the left and the right is empty, extracting the left into a mono file gives you a one-channel deliverable without keeping an unused second channel. It can be played through both sides of an ordinary stereo playback system or placed on a centered track in an editor.
Averaging a voice channel with a silent channel is a different operation: this tool’s Mono mix halves that signal’s sample amplitude. Choose the channel directly when the empty side adds nothing you want. The tool does not automatically restore the level or normalize the export.
If the other side contains useful ambience, music, or a second speaker, save it as well. Keep both at the same timeline start in your editor and decide how to balance them there. Extracting a channel should be a deliberate edit, not a reason to discard audio before listening to it.
Why a stereo-to-mono mix can lose sound
Saturalabs’ free channel tool averages corresponding samples: (left + right) / 2. The Web Audio specification gives this stereo-to-mono downmix rule. It is a predictable combination, not an assessment of which channel sounds better.
Consider three simple examples. Identical channels keep their sample amplitude. A signal paired with silence is halved. Equal opposite-polarity signals sum to zero. Listen to and download the three channel examples above to compare their stereo sources, individual channels, and mono mixes. The original synthesized tones make these relationships exact; in a real recording, the relationship can vary over time and frequency, so the effect may be a thinner voice or a weakened instrument rather than total silence.
Compare the mono preview with each individual channel through the same playback setup. Listen for a wanted sound becoming unusually quiet, hollow, or absent. If the mix is unsuitable, use a useful individual channel, keep the original stereo recording, or investigate the microphone alignment in an editor. Channel extraction alone does not repair the relationship between two microphones.
When channels cannot give you a clean source
In a finished song, vocals, drums, and instruments may share both channels. In an interview, each microphone can pick up the other person and room noise. Splitting left and right gives you control over those existing recordings, but every sound within the selected channel remains.
If you still need one audible source from that mix, test source isolation on a representative excerpt. Use the voice isolator for spoken voice or the stem separation workflow for a named musical part. Saturalabs requests one source per pass and returns that target plus the remaining audio; it does not turn two channels into an automatic four-stem project.
For a vocal-reduced backing track, describe “vocals” and inspect Background audio. For the requested voice or instrument itself, inspect Isolated sound. AI separation uses the product’s separate account and download-credit flow. The channel splitter’s local downloads remain free.
Prepare an output you can trace and reuse
- Keep the original stereo recording. Export working copies rather than replacing your source.
- Check both channels. Listen to the same representative passage on each side.
- Name the choice accurately. Use “left-mono” or “right-mono,” not “isolated voice” unless the recording actually contains only that voice.
- Retain the timeline position. If the input was an excerpt, note its original start time. Place both channel exports at that same time in the edit.
- Review the final context. Check dialogue continuity, stereo placement where relevant, and the actual exported edit.
The free splitter accepts two-channel MP3/WAV recordings up to 50 MiB and five minutes. It exports 48 kHz, 16-bit mono PCM WAV and does not copy tags or preserve compressed bytes. For a longer recording or video, create a manageable audio excerpt in your editor first. The free trimmer and file size calculator help with supported short recordings and planned exports.
The same duration and decoding settings apply to both selected channels, but that is not a promise of sample-for-sample correspondence with a compressed original. Keep the original file, and check alignment after import—especially when returning an excerpt to a video timeline.
Common questions
Can a mono file contain several instruments?
Yes. One channel can contain a complete musical mixture. Channel count does not tell you how many sound sources are present.
Does splitting stereo remove background noise?
Only if choosing a channel omits audio that was confined to the other side. Noise mixed into the selected channel remains. Use the cleanup method guide to choose a separate treatment for that problem.
Are left and right channels separate stems?
They can hold separate source recordings if the recording was arranged that way. In a finished mix they often contain overlapping material. Listen and check the recording setup rather than relying on the channel labels.
Should I convert every stereo file to mono before isolation?
No blanket conversion is needed. First determine whether a particular channel already serves your goal. A mono combination can change wanted audio, and reducing channel count is not evidence of better separation.