Mono-Compatibility Checks for Social Mixes
Mid-side generated music can cancel in mono on TVs and some phones. Run a ten-second fold, read a correlation meter, and collapse width without killing the bed.
A generated bed that sounds huge in headphones can disappear on a TV and on some phones. The usual cause is a stereo trick that does not survive a fold to one channel: mid-side width, a Haas delay, a chorus on the side, a bass part that was not actually centred. A ten-second mono sum plus a correlation meter tells you whether the bed is still there. Collapsing the side a few dB usually fixes it. Replacing the generate is the fix when the side is the track.
Text-to-music models are trying to deliver a finished stereo record. That is the point of the product. It is also why their beds misbehave under a voice on a kitchen speaker.
Why generated beds fail in mono
Stereo is two channels. Mono playback adds them. Anything equal in both channels (the mid) gets louder. Anything opposite in the two channels (the side) shrinks or cancels.
Generated music leans on the side because width reads as "production value" on a stereo meter and in headphones:
- Mid-side enhancement. A lot of modern mastering (human or model) pushes S relative to M. Headphones love it. A mono fold does not.
- Haas / short delay widening. Copy a part, delay it a few milliseconds, pan the copy opposite. In stereo it is a room. In mono it is a comb filter, and the part dips.
- Out-of-phase low end. A bass or pad that is not centred, or that has a stereo chorus on it, loses weight when summed. You feel this as "the bed died" even if the hats are still there.
- A vocal in the generate. A sung hook sitting wide, or a stereo doubled lead, fights your VO in stereo and then still fights it in mono, unless you split the song and use the instrumental.
Playback that folds toward mono is not rare: a single phone speaker, many TVs, a bathroom Bluetooth speaker, one earbud. You do not need a device census. You need the fold on your file.
Native-audio video can add a second version of this: a source that pans with the camera. Spatial audio that pans with the camera is still two channels. Check the fold, because the espresso machine that moved from left to right still has to exist when left and right become one.
The ten-second fold and the correlation meter
Two checks, in that order. The fold is the one your ear trusts. The meter is the one that names the failure.
1. Sum to mono for ten seconds, with the VO playing.
In the DAW, fold the output (a utility plugin, a mono switch on the master, or pan both channels to centre on a summing bus). Do not solo the bed. The question is whether the spot still has a bed and a voice, not whether the instrumental is still impressive.
Listen specifically for:
- A hole where the kick or the pad was
- The VO getting thinner (phasey) rather than just quieter
- A sung ghost that got louder relative to the instruments
- FX that were hard-panned (a whoosh only on the right) dropping 6 dB or vanishing
Ten seconds is enough if it includes a chorus and a verse, or a hook and a line of VO. Looping a 2-bar intro that was mixed centred will fake a pass.
2. Read a correlation (goniometer / phase) meter on the bed, then on the printmaster.
A stereo correlation meter reads about +1 when left and right are the same (perfectly mono-compatible), around 0 when they are unrelated, and toward -1 when they are out of phase. You do not need a magic threshold from a forum. You need the needle to stay on the positive side for a bed that must survive a fold, especially in the low end.
| Reading (bed, low-passed if you can) | Meaning | Action |
|---|---|---|
| Mostly +0.5 to +1 | Compatible | Still do the ear fold; meters miss musical holes |
| Hovering near 0 | Very wide, little shared mid | Collapse side until the fold has a kick again |
| Dipping negative, especially on kicks or pads | Out of phase energy | Collapse side, or high-pass the side, or replace the generate |
| Printmaster negative when DX is in | Voice and bed fighting in a weird way | Centre the DX; the voice should be mid, not a stereo effect |
If you have no plugin meter, the fold is still mandatory. The meter names the failure. It does not replace the ear.
Do this on the bed, then on the printmaster. A centred VO can hide a rotten side until a TV folds the whole mix and the pad drops out under the words.
How to collapse width without killing the bed
The wrong fix is "put a mono plugin on the master and call it compatible." That throws away every stereo hat and every useful space, and it will make you turn the bed up until it masks the VO again.
Work in mid-side on the MX (and on FX if a hard pan is the hole).
- Encode M/S. Mid is the sum, side is the difference. Most EQ and imaging plugins have this as a mode.
- High-pass the side. Start around 120-200 Hz on S only. Low-frequency stereo is the cancellation that feels like the bed died. The kick and bass belong in M.
- Lower the side. A few dB, then fold again. You are looking for the point where the mono kick returns and the stereo hats are still hats. If you have to pull S more than something like 6-8 dB, the generate is a width effect with a suggestion of a song. Replace it.
- Do not high-pass the mid to "clean" the VO. That is the voice's body if the bed was already carved, and it is the kick if it was not. Carve 250-500 Hz on the bed in stereo (or in M) as a masking fix, separately. Width is not masking. Masking is the muddy-VO problem.
- Re-centre anything that must survive: VO, kick, bass, the product hit. If a whoosh is a picture cue, keep it a bit off-centre if you like, then confirm the fold still has an event at that frame, even if it is quieter.
Prompt the next generate so you are not fighting a brick of S:
Instrumental only, no vocals, mono-compatible mix, bass and kick centred, no stereo chorus on the low end, modest width, sparse enough to sit under a voiceover.
The music generator will still try to make a record. "Bass centred" and "instrumental only" are the words that change the fold. "Wide, lush, cinematic stereo" are the words that refill the side.
If the track you already like has a singer, split it and use the instrumental as MX. Stem separation is inferred, with bleed. It is still the right first move when the cancellation is a wide doubled vocal you never wanted under the VO.
Adding the music to the video in mix mode does not collapse width. It sets a static level. Do the M/S work in the DAW, bounce a stereo MX that already survives a fold, then attach.
A check that belongs in the export ritual
Put the fold where you already listen, not in a QA doc nobody opens.
- Mix on the chain you trust.
- Fold to mono, ten seconds, VO up. If the bed holes out, collapse S or replace MX.
- Correlation glance on MX and on the printmaster.
- Phone-speaker pass (one driver, in a real room).
- Then export the picture.
If you assemble in Versely first, preview is picture. The editor's 480p preview (preview: true) is free, with a short per-user cooldown, and it will play stereo through whatever speakers the laptop has. That playback is not a mono check. Bounce the audio, fold it, then export.
A mix that only works in headphones is not a social mix. It is a demo of the generate. Collapse the side until the kitchen speaker still has a kick and a sentence, then stop. Width you cannot hear in mono was never going to travel.
FAQ
Does collapsing width make the mix "small"?
On headphones, a little. On the devices that fold, it makes the mix exist. If the spot is a headphone-first film with no VO, you can keep the side. If the spot is a talking-head ad, the voice is the programme and the bed is furniture. Furniture that vanishes in mono is a hole, not a vibe.
Is a correlation reading of 0 a fail?
It is a warning, not an automatic fail. Uncorrelated stereo can still sum without a hole if there is enough mid. Negative correlation, especially in the bass, is the fail. Always confirm with the ear fold. A meter that never leaves +1 means you are already mono, which is compatible and also a bit lifeless. That is a taste choice, not a phase problem.
Will Versely fold to mono for me?
No. Generate, attach, and export are stereo paths. The fold is a DAW or NLE switch, or a bounce of a mono sum you listen to once. Do not ask the agent to "make it mono-compatible" and expect an M/S processor. Ask for a sparser, instrumental, centred-bass generate, then fold what you get.
What if the hole is an FX whoosh, not the music?
Hard-panned FX lose about 6 dB in a simple sum, or they vanish if they were polarity tricks. Keep picture-critical events closer to centre, or add a quieter centred copy so the fold still has a hit. Do not widen a sync effect to make the mix "bigger." The picture is already the location.