Mono, Stereo, 5.1, and What a Downmix Does

We isolated each channel of a synthetic 5.1 file and ran it through FileFlip's own downmix path, the same one Channels: Stereo triggers on a real one. The center channel, where dialogue usually sits, lands 3 dB quieter than the front left and right once the mix is folded down. The LFE channel doesn't get quieter at all. It disappears.

Should you convert to mono, stereo, or leave 5.1 alone?

Use mono when:

• The source is voice only: a podcast, a voicemail, a lecture recording, anything with nothing on the sides worth keeping.
• The target is an uncompressed or fixed-quality format like WAV or FLAC, which is where mono actually saves space. Convert WAV to MP3 instead and Channels stops affecting the file size at all; see "Does converting to mono actually halve the file size?" below.

Use stereo when:

• The source has any real channel separation: music, a movie soundtrack, anything meant for headphones or two speakers.
• You're not sure. It's FileFlip's default for anything that started with more than one channel.

Leave 5.1 alone when:

• The file is headed to a home theater receiver, a broadcast pipeline, or a video container built to carry six discrete channels. That equipment can fold the mix down better than a fixed formula baked into a file ahead of time.

Convert AC-3 to M4A when:

• The destination is a browser, a phone, or a web player. Almost nothing outside a TV or an AV receiver decodes Dolby's codec natively, and AAC plays everywhere. Convert AC-3 to M4A and see "Where's the channels control in FileFlip?" below for what happens to the six channels on the way.

Channel layouts and formats at a glance

The three formats this post covers, side by side:

Does converting to mono actually halve the file size?

Only in an uncompressed file. We converted our 7.47-second speech WAV to mono and it dropped from 1,316,926 bytes to 658,502 bytes, 50.0% smaller, exactly what dropping from two channels of samples to one predicts.

MP3 and AC-3 don't follow that pattern, because a bitrate is a total budget, not a per-channel one. We encoded the same file to MP3 at five bitrates, once as mono and once as stereo, and every pair landed on the identical byte count:

AC-3 matched that at every bitrate we tried too, from 128 through 448 kbps. Setting Channels to Mono on a lossy target doesn't shrink the file. It spends the same total bits on one channel instead of splitting them across two, so quality per channel goes up and the byte count doesn't move.

FLAC, a lossless format, sits between the two. Our mono file came out 236,893 bytes against 237,118 bytes stereo, a 0.1% gap, far short of the WAV result. This corpus's left and right channels are identical to start with, so FLAC's own stereo coding had already squeezed out most of the redundancy that going mono removes. A source with real stereo separation would likely show a bigger drop, though we don't have one to measure here.

What does a 5.1-to-stereo downmix do to dialogue?

Dialogue almost always lives in the center channel, so what happens to that one channel in a downmix matters more than what happens to the other five. We built a synthetic 5.1 file with a single test tone isolated in one channel at a time, silence everywhere else, and ran each through ffmpeg -ac 2, the exact command FileFlip's Channels: Stereo control issues.

The center channel lands 3.0 dB quieter than the front left and right once it's folded in alongside them, on top of the 7.7 dB every channel loses so six channels playing at once don't clip when they're summed into two. That 3 dB gap isn't a rounding error. It's FFmpeg's standard center and surround mix level, a factor of 1 over the square root of 2, stacked on the same headroom cut every channel takes.

The arithmetic explains the headroom too: 0.414 (front) plus 0.293 (center) plus 0.293 (surround) adds up to exactly 1.0, which is why the front channels get pulled down as far as they do. The whole matrix is built so a moment where all six channels hit full volume at once still can't clip the two they land in.

The LFE channel doesn't get pulled down. It gets dropped. Dolby's own description of the format says the ".1" channel needs about a tenth the bandwidth of the other five because it only carries bass, and FFmpeg's default downmix doesn't fold that bass into the stereo mix at all.

Where does AC-3 come from?

AC-3 is Dolby Digital's own name for its codec, and it's the format behind most of the 5.1 files you'll run into. Dolby Digital was first used theatrically on Batman Returns in 1992, and it became the default 5.1 audio track on DVD and most broadcast TV afterward.

Dolby's own numbers put 5.1 Dolby Digital on DVD at 384 or 448 kbps, and Blu-ray discs up to 640 kbps. More channels don't automatically cost more bits with a bitrate-targeted codec, as the mono/stereo table above shows, but six real channels need more room to sound clean than two do, which is why the standard bitrate climbs with the channel count instead of staying fixed.

Where's the channels control in FileFlip?

Under Advanced options on any audio conversion. It only offers two choices, Mono and Stereo. There's no separate 5.1 option and no control for how loud the center channel should land in the downmix; picking Stereo just runs the -ac 2 conversion measured above.

Leave Channels untouched and FileFlip keeps whatever channel count the source had. We ran a synthetic 5.1 AC-3 file through a conversion to M4A with Channels left alone, and the output AAC file kept all six channels, at 149,618 bytes. Setting Channels to Stereo on the same file produced a 91,468-byte, two-channel file, using the exact downmix measured above.

So the arbitrary part of folding 5.1 down to stereo isn't which channels get picked. FileFlip, like most encoders, never asks where dialogue should sit in the result. It applies the one fixed curve, every time.

Common questions

Does more channels always mean a bigger file?

Only for an uncompressed format like WAV, where every channel is its own stream of samples. A lossy format with a fixed bitrate spreads the same total bits across however many channels you give it, so more channels mean lower quality per channel, not a bigger file.

Can FileFlip convert 5.1 audio straight to mono?

Yes. Set Channels to Mono on a 5.1 source and FFmpeg mixes all six channels down to one, using the same kind of fixed matrix measured above rather than picking a single channel and discarding the rest.

Why convert AC-3 to M4A instead of MP3?

Because FFmpeg's MP3 encoder can't hold more than two channels at all. Feed it our six-channel test file with Channels left alone and it downmixes to stereo automatically, whether you asked for that or not. AAC, the codec inside an M4A file, has no such ceiling; see MP3, AAC, Opus, Vorbis and FLAC for how it compares to MP3 more generally.

Is the downmix the same for every file?

Yes, on FileFlip and on FFmpeg generally. It's a fixed formula applied to the channel layout, not something tuned per file, which is why the measurement above is reproducible on any 5.1 source.

Convert between them

FileFlip converts AC-3 to M4A and WAV to MP3 entirely in your browser, using a WebAssembly build of FFmpeg. The file never leaves your machine: nothing is uploaded, there's no account, and there's no queue. Open Advanced options to set Channels to Mono or Stereo, matching the behavior measured above.

For the format details, see the AC3, WAV and MP3 format references.

How we measured this

• Corpus: the same 7.47-second speech WAV used across our audio posts, 44.1 kHz, 16-bit, stereo, fetched with pnpm blog:corpora. Its left and right channels are identical, confirmed by extracting each with FFmpeg's pan filter and comparing them byte for byte.
• Mono/stereo size table: ffmpeg -i speech.wav -ac 1 -b:a <n>k out.mp3 against the same command with -ac 2, for MP3 at 96 through 320 kbps and AC-3 at 128 through 448 kbps. FFmpeg 8.1.2 on macOS, the library FileFlip runs as WebAssembly.
• Downmix gain table: built a synthetic 5.1 WAV with FFmpeg's sine, anullsrc and join filters, one -12 dBFS, 1 kHz tone isolated in a single channel of the FL+FR+FC+LFE+BL+BR layout at a time, silence in the rest. Downmixed each with ffmpeg -ac 2, no other filter, the same flag FileFlip's own encoders receive from the Channels control. Measured the RMS level of each output channel, isolated with pan, using FFmpeg's astats filter.
• Channel-count test: built a second 5.1 WAV with real, non-silent content in all six channels, encoded it to AC-3 at 448 kbps, then transcoded that file to AAC/M4A twice, once with Channels left unset and once set to Stereo.
• Caveat: the FLAC and downmix numbers both come from one synthesized corpus, and its left and right channels are identical. A file with genuine stereo separation, or six channels of real surround content instead of isolated test tones, would move the FLAC gap and the LFE table's absolute levels. It would not change the mono/stereo bitrate-matching result or the downmix formula itself, since both are properties of the encoder, not the content.

The commands above are copy-pasteable and reproduce every number in this post; there's no bundled script for multichannel audio yet, unlike the single-track benchmarks pnpm blog:bench-audio runs.