How to isolate a channel from stereo audio for editing
- Step 1Confirm the file is a 2-channel stereo — Isolation works on stereo. After you add the file, check the file row reads
2ch(notmono). Only a genuine two-channel file holds two separable sources. - Step 2Drop the stereo session file in — Add a stereo
.wav,.mp3,.m4a,.flac,.ogg, or.opus. FFmpeg 8.1 runs in WebAssembly in the page — your stems never leave the device. - Step 3Run the split into stems — The tool runs
channelsplit=channel_layout=stereoand produces two mono WAV stems. No parameters to set. - Step 4Download both stems — File
-1is the left-channel stem,-2is the right-channel stem. Use Download all to get both; they stay sample-aligned. - Step 5Clean each stem independently — Run the noisy stem through ai-noise-reducer, the quiet voice through speech-leveler, and shape tone per stem with voice-eq — no cross-contamination.
- Step 6Re-assemble the polished stems — Bring the cleaned stems back together with audio-merger, then set final loudness with loudness-normalizer.
Channel isolation vs source separation
This tool isolates channels. It does not isolate instruments mixed across channels.
| You want to isolate… | Works here? | Why |
|---|---|---|
| Two mics on a dual-mono file | Yes | Each mic is one channel — isolation = its own stem |
| Host vs guest on a two-track recording | Yes | Speakers are hard-split across L and R |
| A drum bus vs a vocal in a stereo song | No | Both are mixed into both channels — needs AI source separation, not channel split |
| A hard-panned instrument | Partly | You get the whole side it's on, including anything else panned there |
| One channel for a mono check | Yes | Isolate the channel and audition it alone |
Stem output facts
From the processing code — fixed behaviour, no options panel.
| Property | Value |
|---|---|
| Filter | channelsplit=channel_layout=stereo (FFmpeg 8.1 WASM) |
| Input required | Stereo (exactly 2 channels) |
| Stems out | Two mono files — -1 left, -2 right |
| Format | 16-bit PCM WAV, original sample rate |
| Alignment | Sample-accurate — stems stay in sync |
| Upload | None — in-browser |
Tier limits
Audio family. Duration cap is per file, separate from size.
| Tier | Max size | Max duration | Files/run |
|---|---|---|---|
| Free | 50 MB | 30 min | 1 |
| Pro | 200 MB | 120 min | 10 |
| Pro + Media | 100 GB | Unlimited | 100 |
| Developer | 100 GB | Unlimited | Unlimited |
Cookbook
Stem-prep recipes for per-channel cleanup. Left stem is file -1, right stem is -2; they stay sample-aligned.
Noise-reduce only the noisy mic
A two-mic interview where the guest's mic (right) picked up a fan but the host's (left) is clean. Isolate, then de-noise only the noisy stem.
Input: ep.wav (2ch, host=L clean, guest=R fan noise) Stems: ep-channel.wav-1 (host, clean) ep-channel.wav-2 (guest, noisy) ai-noise-reducer on -2 only, then merge. The clean host stem is untouched.
EQ each voice to its own profile
A deep-voiced host and a brighter guest. Isolate so you can apply a warm profile to one and a bright profile to the other.
Stems: -1 host -> voice-eq profile: warm -2 guest -> voice-eq profile: bright Then merge -> tonally matched two-voice track.
Level a loud guest without pumping the host
Per-channel leveling avoids the classic problem of a single compressor reacting to the loudest source across both speakers.
Stems: -1 host -> speech-leveler (preset: interview) -2 guest -> speech-leveler (preset: interview) Each stem leveled independently, then merged.
Re-align and re-merge after editing
Because the split is sample-accurate, the edited stems drop back into a merge with no drift.
After cleanup: -1 host-clean.wav (edited) -2 guest-clean.wav (edited) audio-merger -> single track; stems were never time-shifted.
What it can't do: separate a song's instruments
People expect 'isolate a channel' to extract drums or vocals from a finished stereo song. Channel splitting can't — the instruments are mixed into both channels.
Input: finished-song.wav (2ch, full mix) Isolate L -> still the full mix (just the left side) Isolate R -> still the full mix (just the right side) For true stem separation you need an AI source-separation tool; channel split only deinterleaves L and R.
Edge cases and what actually happens
Trying to isolate an instrument from a stereo mix
Not supported hereChannel isolation only separates the left and right channels. In a finished stereo song, every instrument is mixed into both channels, so neither isolated channel is a clean instrument stem. That requires AI source separation, which this tool does not perform.
Input is mono
Rejected (requires stereo)A mono file holds a single channel — there are no separate sources to isolate. The tool throws Channel splitter requires a stereo input.
Stems sound identical
By designIf the source is dual-mono (the same audio on both channels), both stems will be identical. The split faithfully copied each channel — there simply weren't two different sources to begin with.
Hard-panned source has bleed from others
By designIsolating a channel gives you everything panned to that side, not just one source. If other elements were also panned there, they appear in that stem. Isolation is channel-level, not source-level.
Multichannel session file (more than 2 channels)
Rejected (requires stereo)Only exactly 2 channels are accepted. A multichannel stem export is rejected with the stereo-required error; this tool does not deinterleave 5.1/7.1 into per-channel stems.
Stems come out as WAV, not the input format
ExpectedStems are always 16-bit PCM WAV — the best starting point for further DSP and re-merge. There's no format option. Re-encode later with wav-to-mp3 or wav-to-flac.
Stems drift out of sync after editing
Preserved (not from the split)The split itself introduces zero offset — both stems start at the same frame. Any drift comes from an edit that changed one stem's length (a trim or time-stretch). Keep lengths matched, or align in your DAW before merging.
Long session over the duration cap
413 tier limitFree caps duration at 30 minutes per file regardless of size. Long session files need Pro (120 min) or Pro + Media (unlimited). Duration and size limits are enforced separately.
Frequently asked questions
Can I isolate one instrument from a stereo song?
No. Channel isolation separates the left and right channels, but a finished song mixes every instrument into both channels. Neither isolated channel is a clean instrument stem. Genuine instrument separation needs an AI source-separation process, which this tool does not do.
What do I get out — and in what format?
Two mono stems as 16-bit PCM WAV at the original sample rate: file -1 is the left channel, -2 is the right. WAV is chosen so the stems are a lossless base for de-noising, EQ, and re-merge.
Do the stems stay in sync?
Yes. The split is sample-accurate and applies no offset, so both stems start on the same frame and re-merge with zero drift — as long as your edits don't change one stem's length.
Why would I isolate channels before editing?
So you can process each source independently. De-noise only the noisy mic, compress only the loud speaker, EQ each voice to its own profile — none of which is possible while both sources share one stereo file.
Can I isolate a hard-panned instrument from a rough mix?
Partly. You'll get the entire side that instrument is panned to, including anything else panned the same way. If it's the only thing hard-panned there, that's effectively an isolated stem; otherwise it has company.
Does it work on a mono recording?
No — a mono file has one channel and no separate sources, so it's rejected with Channel splitter requires a stereo input. For width on a mono voice, use mono-to-stereo.
How do I put the stems back together after cleanup?
Use audio-merger to recombine the edited stems, then loudness-normalizer to set the final delivery level.
Will isolating reduce quality?
No. The deinterleave is sample-exact and the stems are lossless WAV. If the source was lossy (e.g. MP3), those existing artefacts carry over, but the split adds no new loss.
Are my session stems uploaded?
No. FFmpeg 8.1 runs in your browser via WebAssembly, so unreleased music or confidential interview stems stay on your device.
Which input formats can I use?
WAV, MP3, M4A/AAC, FLAC, OGG, and Opus, provided the file is stereo. FFmpeg decodes the container before splitting.
Can I run a whole batch of session files?
Per-run file counts follow your tier (Free 1, Pro 10, Pro + Media 100, Developer unlimited). This tool isolates one stereo file at a time; run them in sequence within your allowance.
How long a session can I split into stems?
Free allows up to 30 minutes / 50 MB per file; Pro 120 minutes / 200 MB; Pro + Media removes the duration cap (100 GB). Uncompressed session WAVs are large, so longer files generally need a higher tier.
Privacy first
Every JAD Audio tool runs entirely in your browser via FFmpeg (WebAssembly) and RNNoise. Your audio files never leave your device — verified by zero outbound network requests during processing.