How to split a stereo file into two mono files
- Step 1Confirm your file is dual-mono stereo — This tool splits a stereo (2-channel) file. After you add the file, the tool reads its channel count via the browser's audio decoder and shows
monoor2chin the file row. If it saysmono, there is nothing to split — the two speakers were already summed together and cannot be separated. - Step 2Drop the stereo recording onto the tool — Drag in your
.wav,.mp3,.m4a,.flac,.ogg, or.opusexport. FFmpeg 8.1 runs inside the page in WebAssembly, so the interview audio stays on your machine — nothing is uploaded. - Step 3Run the split — There are no settings to choose — the tool is a fixed-behaviour splitter. It executes
channelsplit=channel_layout=stereo, mapping[left]to the first file and[right]to the second. - Step 4Download both mono files — The result panel lists two files. Use Download all to save both, or the per-file links to grab just one. Files are named from your source with a
-channelsuffix and a-1/-2index (file-1is the left channel,-2is the right). - Step 5Process each voice independently — Now that the host and guest are in separate mono files, level each one with speech-leveler, remove room noise with ai-noise-reducer, or shape tone with voice-eq — without one speaker's bleed fighting the other.
- Step 6Re-combine into the final mix — When both voices are clean, bring them back together with audio-merger, then set a single delivery loudness with loudness-normalizer before publishing.
What the splitter actually does
Fixed behaviour — there are no per-run options to configure. These are the real input/output facts from the processing code.
| Property | Behaviour | Why it matters |
|---|---|---|
| Filter used | channelsplit=channel_layout=stereo (FFmpeg 8.1, WASM) | True per-channel deinterleave — file 1 = exact left samples, file 2 = exact right samples |
| Required input | Exactly 2 channels (stereo) | Mono or multichannel (5.1, etc.) is rejected with Channel splitter requires a stereo input. |
| Output count | Two mono files | -1 is the left channel, -2 is the right channel |
| Output format | WAV, 16-bit PCM (pcm_s16le), original sample rate, mono | No format picker in the UI — output is always lossless WAV regardless of input type |
| Naming | <source>-channel.wav then a -1 / -2 download index | Predictable, so a batch of episodes stays sorted by speaker |
| Processing location | In-browser WebAssembly; no upload | Embargoed or confidential interviews never touch a server |
Tier limits for stereo splitting
Audio-family limits. The duration cap is per file and is separate from the size cap — a long full-quality WAV can hit the duration limit while well under the byte limit, or vice-versa.
| Tier | Max file size | Max duration / file | Files per run |
|---|---|---|---|
| Free | 50 MB | 30 min | 1 |
| Pro | 200 MB | 120 min | 10 |
| Pro + Media | 100 GB | Unlimited | 100 |
| Developer | 100 GB | Unlimited | Unlimited |
Rough stereo-WAV size guide
Uncompressed 16-bit PCM stereo. Use this to judge whether your raw interview WAV fits the Free 50 MB ceiling or needs Pro.
| Sample rate | Per minute (stereo WAV) | 30 min | Fits Free 50 MB? |
|---|---|---|---|
| 44.1 kHz | ~10.1 MB | ~303 MB | No — use a compressed export or Pro |
| 48 kHz | ~11.0 MB | ~330 MB | No — use a compressed export or Pro |
| A 192 kbps stereo MP3 | ~1.4 MB | ~42 MB | Yes (just under) |
Cookbook
Real dual-mic splitting situations, with the exact before/after. The two outputs are always the left channel first (-1) and the right channel second (-2).
Two-track Riverside interview into host + guest
Riverside's combined export puts the host on the left and the guest on the right. Splitting gives you two clean mono voices to process separately before mixing.
Input: interview-ep42.wav (2ch / 48 kHz, host=L, guest=R) FFmpeg (in-browser): channelsplit=channel_layout=stereo -> [left] [right] Outputs: interview-ep42-channel.wav-1 (mono, host) interview-ep42-channel.wav-2 (mono, guest) Next: speech-leveler on each, then audio-merger to recombine.
Skype dual-track recording where one side is louder
Skype's two-track recording often captures one participant hotter than the other. Split first, then level each voice to the same target so neither buries the other.
Input: call-2026-06-10.m4a (2ch, you=L quiet, them=R loud) After split: call-2026-06-10-channel.wav-1 (you) call-2026-06-10-channel.wav-2 (them) Level each with speech-leveler (preset: interview), then merge -> balanced two-voice track.
Field recorder set to dual-mono (lav + boom)
A recorder configured for dual-mono writes the lapel mic to one channel and the shotgun/boom to the other on a single stereo file. Split to pick or blend the better mic per moment.
Input: scene7_take3.wav (2ch, lav=L, boom=R) Outputs: scene7_take3-channel.wav-1 (lav mic) scene7_take3-channel.wav-2 (boom mic) Use whichever mic is cleaner; both are full-quality WAV.
MP3 episode in, lossless WAV stems out
Even when the source is a lossy MP3, the split itself does not re-encode to MP3 — both channel files come out as 16-bit PCM WAV, so no second generation of MP3 loss is added before editing.
Input: draft.mp3 (2ch, 192 kbps) Outputs (note: WAV, not MP3): draft-channel.wav-1 draft-channel.wav-2 The MP3's existing artefacts are preserved as-is; the splitter adds no new lossy stage. Re-encode to MP3 later if you want with wav-to-mp3.
Mono file dropped in by mistake
If the recorder summed both mics to a single mono channel, there is nothing to separate. The tool detects this and refuses rather than producing two identical files.
Input: voice-memo.m4a (1ch / mono) Result: Error: "Channel splitter requires a stereo input." Fix: there is no fix after the fact — a true mono recording cannot be unmixed into two speakers. For perceived width on a mono voice, use mono-to-stereo instead.
Edge cases and what actually happens
Input is mono
Rejected (requires stereo)The tool checks the decoded channel count and throws Channel splitter requires a stereo input. for any mono file. A genuinely mono recording has only one channel of audio — there is nothing to separate, and a split would just duplicate it. For adding width to a mono voice, see mono-to-stereo.
Input has more than 2 channels (5.1, 7.1)
Rejected (requires stereo)The guard is exactly 2 channels, so a 5.1 or 7.1 file is also rejected with the same stereo-required error. This tool is purpose-built for L/R dual-mono. Multichannel deinterleaving is out of scope here.
File row shows '2ch' but both speakers come out on both files
By designThe split is faithful to the source. If your 'two-track' export was actually mixed (both voices summed into a correlated stereo image rather than hard-panned per channel), each output will contain both voices because that is what each channel genuinely holds. Splitting cannot un-mix audio that was already blended — only true dual-mono separates cleanly.
Joint-stereo MP3 source
SupportedFFmpeg fully decodes joint-stereo (mid/side) MP3 back to discrete L/R PCM before the split, so a joint-stereo MP3 separates into the correct two channels. The split itself outputs WAV, so the joint-stereo encoding is gone from the outputs.
Output files are WAV when you expected MP3
ExpectedThis splitter has no output-format control — it always writes 16-bit PCM WAV. That is intentional: starting from lossless avoids stacking a fresh MP3 generation on top of your edits. If you need MP3 deliverables, run each split file through wav-to-mp3 afterwards.
Large full-quality WAV exceeds the Free duration cap
413 tier limitFree tier allows 30 minutes per file; a long interview WAV will hit that even when it is under a size limit. The duration cap is enforced separately from the byte cap. Pro raises both to 120 minutes / 200 MB; Pro + Media removes the duration cap entirely.
Browser cannot decode the container for metadata
PreservedChannel detection uses the browser's Web Audio decoder. For some FLAC/Opus files on Safari, decode can fail and the tool falls back to a default channel assumption — in that case verify your file really is stereo before relying on the result. FFmpeg itself still decodes the audio correctly for the actual split.
Very short clip (under a second)
SupportedThere is no minimum length. A sub-second stereo sting or test tone splits the same way as a full episode — two mono WAV files, one per channel.
Expecting a single ZIP of both channels
By designThe outputs are delivered as two separate downloadable files (via 'Download all' or per-file links), not bundled into one ZIP. Save both before leaving the page.
Frequently asked questions
Which output file is the left channel and which is the right?
File -1 is the left channel and file -2 is the right channel. The tool maps FFmpeg's [left] output to the first file and [right] to the second, so for a host-left/guest-right export, -1 is the host and -2 is the guest.
What format are the two output files?
Both are mono WAV files — 16-bit PCM (pcm_s16le) at the original sample rate. There is no format chooser; the splitter always writes WAV so you start editing from a lossless source. Convert to MP3 afterwards with wav-to-mp3 if needed.
Can I split a mono file into two?
No. A mono file has a single channel, so there is nothing to separate — the tool rejects it with Channel splitter requires a stereo input. If you want a mono voice to play out of both stereo sides (or to add perceived width), use mono-to-stereo.
Does splitting reduce audio quality?
The split is a sample-exact channel deinterleave and outputs lossless WAV, so it adds no quality loss of its own. If your source was a lossy MP3, the existing MP3 artefacts carry over (they are baked into the audio), but no new lossy stage is introduced.
Why does each output still contain both speakers?
Because the source channels themselves contain both speakers. Splitting copies each channel exactly — it cannot un-blend audio. Clean separation only happens when the recording is true dual-mono (one speaker hard-panned to each channel), as Zencastr, Riverside two-track, and Skype two-track exports usually are.
How do I recombine the two mono files after editing?
Use audio-merger to bring the two cleaned voices back into one track. If instead you want to rebuild a stereo file from a single processed voice, mono-to-stereo duplicates or widens it.
Is my recording uploaded anywhere?
No. Processing runs entirely in your browser via FFmpeg 8.1 compiled to WebAssembly. The audio never leaves your device, which is why this is safe for embargoed, legal, or otherwise confidential interviews.
What input formats can I drop in?
Common audio containers — WAV, MP3, M4A/AAC, FLAC, OGG, and Opus — as long as the file is stereo. FFmpeg decodes the container to PCM internally, then splits and writes WAV.
How long an interview can I split?
Free tier allows up to 30 minutes and 50 MB per file. A 30-minute full-quality stereo WAV is around 300 MB, so for long uncompressed episodes you'll want Pro (120 min / 200 MB) or Pro + Media (unlimited duration / 100 GB). A compressed MP3 episode usually fits Free comfortably.
Can I split several episodes at once?
On the audio family, Free handles one file per run, Pro up to 10, Pro + Media up to 100, and Developer is unlimited. This particular tool takes one stereo file at a time; for batch throughput you'd run them in sequence within your tier's per-run allowance.
Do the channels stay in sync after splitting?
Yes. Both outputs come from the same source frames at the same sample rate with no offset applied, so the two mono files line up sample-for-sample when you stack them in a DAW or feed them to audio-merger.
The file row says my file is mono but I recorded in stereo — what happened?
Either the recorder summed both mics to mono (common on phone voice memos and some default app settings), or the browser could not decode the channel layout. If it truly recorded mono, the speakers cannot be separated after the fact — re-record with a true two-track/dual-mono setting next time.
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.