Immersive mixing best practices

Ten non-negotiable rules for mixing music in 3D, with checks you can run on every mix.

This page defines the ten rules that every immersive music mix should satisfy before release, whether you deliver Dolby Atmos, Eclipsa Audio or a binaural version. Each rule has a definition, a rationale and a check you can run on every mix. The page closes with a pre-release checklist and a structured critical listening procedure.

It is intended for engineers, producers and students who want a repeatable standard. An immersive master is rendered live to whatever the listener has, from a 7.1.4 speaker room to a soundbar, headphones or a stereo fold-down on a phone, so small mistakes carry further than in stereo.

Rule 1: Start from great sources

Definition. The mix starts from a finished, approved stereo mix and from clean, aligned stems.

Rationale. Clipping, noise, phase issues or missing parts in the stems are heard in every speaker, often more clearly than in stereo, because each part is more exposed.

  • Use the finished stereo mix as the creative reference, so the immersive version feels like the same song.
  • Bounce every stem from the same start point and to the same length. Misaligned stems cause flamming, phasing and timing problems.
  • Leave master bus processing off the stems. Limiting and bus compression belong on the immersive mix.
  • Leave headroom below 0 dBFS and avoid clipping.
  • Decide whether effects are printed or dry, and label the stems. Printed reverbs and delays can conflict with the new space.

If you only have a stereo file, Spatial9 can separate it into stems with AI. See Exporting stems.

Check. Confirm that all stems share one start point and length, none clips, no master bus processing is printed and printed effects are labeled.

Rule 2: Calibrate your monitoring or use a trusted reference

Definition. Every monitoring path used for spatial decisions is calibrated for distance, level and angle, or you use a trusted binaural reference.

Rationale. If one speaker is louder or closer than the others, every placement leans towards it. Because of the precedence effect described by Helmut Haas in 1951, a sound that arrives a few milliseconds earlier from one speaker pulls the image towards it.

ParameterRequirement
DistanceSame for every main speaker, or delayed to match
LevelMatched at the listening position using pink noise and a meter or measurement microphone
Front left and rightAbout 30 degrees
Side surroundsAbout 90 to 110 degrees
Rear surroundsAbout 135 to 150 degrees
Height speakersAbout 30 to 55 degrees above ear level

Without a speaker room, mix with a binaural render on good headphones and the Spatial9 virtual 7.1.4 speaker room, then confirm on real speakers before release if you can.

Check. Verify time alignment and matched levels at the mix position, or schedule a speaker confirmation pass for a headphone mix.

Rule 3: Anchor the spine of the song

Definition. The lead vocal, kick, snare and bass sit at the front centre, or very close to it, in every section.

Rationale. The front centre survives every kind of playback, from a full speaker room to a phone speaker. Research by A. W. Mills in 1958 showed that the ability to tell two directions apart is sharpest straight ahead, at around one degree, and much weaker to the sides and behind.

Check. Solo the four spine elements in each section and confirm that each stays at or near the front centre.

Rule 4: Use bed and objects on purpose

Definition. In object-based formats such as Dolby Atmos and Eclipsa Audio, the bed is a set of fixed speaker channels, such as 7.1.2, and objects are individual sources with position information that the playback system renders to the listener's speakers or headphones. Each part is assigned to one or the other deliberately.

Rationale. A Dolby Atmos master can hold up to 128 inputs, commonly a 10-channel 7.1.2 bed plus up to 118 objects. Many objects make a mix harder to manage, can increase processing on playback devices and often sound less cohesive than a well-built bed.

PartUsually best asReason
Drum kit and room micsBedWide, stable and glued together
Pads, keys and ambienceBedTexture without a need for precision
Main reverbsBedDiffuse space across channels
Lead vocalObject or centre bed channelPrecise and centred on every render
Solo instrument or hookObjectClear, focused position
Effects and ear candyObjectMoves or appears in specific places
Height accentsObjectLifted only for chosen moments

Check. For every object, state why it needs precision or movement. Move parts without such a reason into the bed.

Rule 5: Use height with restraint

Definition. The height layer stays lighter than the main layer and carries mainly air, reverbs, shimmer and harmonies.

Rationale. Listeners locate sounds above them less precisely than sounds in front, so a crowded height layer becomes a blur. Headphones and soundbars also render height less convincingly than a speaker room. A full instrument overhead is powerful because it is rare.

Check. Mute the height layer, then restore it. The main layer must carry the song alone, and the height layer must sound lighter.

Rule 6: Treat LFE as an effect

Definition. The low-frequency effects channel, called LFE or the ".1", is a separate, band-limited channel that carries content up to 120 Hz and plays back 10 dB louder than the main channels. It adds weight; it does not carry the bass of the song.

Rationale. Many headphone and stereo renders reduce or ignore LFE, so bass that exists only there can disappear. The playback system routes low frequencies from the main channels to a subwoofer through bass management when needed. Low frequencies are hard to locate, and moving them weakens the groove.

  • Keep bass and kick in the main channels, focused at the front centre.
  • Use LFE for extra weight on drops, hits and sub effects.
  • Never put the only copy of the bass in LFE.
  • Do not move low bass around the room.

Check. Mute LFE and confirm that bass and kick remain fully present and that no low bass source carries movement.

Rule 7: Move sounds only for a musical reason

Definition. Every movement is tied to a moment in the music, such as a fill, a transition, a lyric or a build.

Rationale. The brain is wired to notice sounds that change position. Constant or random movement keeps the listener on alert and is one of the fastest routes to listener fatigue.

  • Let objects land and rest. A gesture that arrives and stops sounds musical; endless motion sounds like a demo.
  • Limit how many things move at once. Two moving objects is often enough.

Check. List every movement with the musical event that justifies it, remove any movement without one and count simultaneous moving objects in the busiest section.

Rule 8: Hit loudness and true peak targets

Definition. The full immersive render meets the delivery targets for integrated loudness and true peak.

Rationale. Dolby recommends delivering Dolby Atmos music at around -18 LUFS integrated with true peaks no higher than -1 dBTP, while stereo streaming masters often sit closer to -14 LUFS or louder. Immersive mixes depend on transients and quiet moments to sound open, and hard limiting makes them sound flat and tiring.

  • Measure with a loudness meter based on ITU-R BS.1770 on the full immersive output, not on one channel or stem.
  • Watch true peak on every channel and object, because inter-sample peaks can clip on playback even when sample peaks look safe.

Check. Confirm both targets on the full render and verify that no limiter is applied only to gain loudness.

Rule 9: Check translation on every kind of playback

Definition. A mix is finished only when it works on the systems the audience actually uses: headphones, earbuds, soundbars, cars and phones.

Rationale. Most listeners never hear a mix on a 7.1.4 system, and each playback type exposes different faults.

PlaybackWhat to check
Binaural headphonesImage outside the head, vocal centred, highs not harsh
Apple MusicUses its own headphone renderer; check there if it is a key platform
5.1 and soundbarsHeight folds down cleanly and nothing important disappears
Stereo fold-downVocal and groove stay strong
Earbuds and phones, low volumeBalance holds, since bass and treble sensitivity changes with level (ISO 226)
Bed summedNo loss of energy or thinning, confirming phase and mono compatibility

Check. Listen on every row of the table and log issues before export.

Rule 10: Deliver clean, correctly formatted masters

Definition. Every deliverable is in the format the destination requires, with clean, consistent files.

Rationale. A mix delivered in the wrong format, at the wrong sample rate or without a required version can be rejected by a distributor or played back incorrectly.

DeliverableRequirement
Dolby AtmosADM BWF at 48 kHz and 24-bit
Eclipsa AudioIAMF for YouTube and supported devices
StereoA stereo master, still needed by many platforms
BinauralFor previews and social media
All filesClear names, same length, same start point

Spatial9 exports Dolby Atmos (ADM), Eclipsa Audio (IAMF), binaural and multichannel versions from the same session. See ADM BWF and Eclipsa Audio and IAMF.

Check. Verify format, sample rate and bit depth per file, confirm every version exists and compare lengths and start points.

Pre-release checklist

Run these twelve checks before every release. If any check fails, fix it before you export.

AreaCheck
MixLead vocal, kick, snare and bass are anchored at the front centre
MixThe height layer is lighter than the main layer
MixEvery movement is tied to a moment in the music
MixLFE adds weight but never holds the only copy of the bass
MeasurementIntegrated loudness is near -18 LUFS for Dolby Atmos music
MeasurementTrue peak is no higher than -1 dBTP
MeasurementNo bed channel or object clips
MeasurementThe bed is checked for phase and mono compatibility
FilesThe Dolby Atmos master is ADM BWF at 48 kHz and 24-bit
FilesAn IAMF version is ready for Eclipsa Audio and YouTube
FilesBinaural and stereo versions are exported
FilesEvery file has a clear name, same length and same start point

Critical listening procedure

Critical listening is focused, structured listening with one question per pass and written notes. A fixed protocol lets you compare mixes fairly and track revisions. It also limits bias: hearing adapts quickly, so a wide sound can start to feel normal, and a sound you expect above you can seem higher because you expect it.

Vocabulary

TermMeaningRecord as
PositionDirection on the horizontal planeClock face: 12 front, 3 right, 6 behind, 9 left
WidthSpace a single source occupiesNarrow, medium or wide
HeightAbove or below ear levelLow, ear level, high or overhead
DistancePerceived nearness; reverb and darker tone also push sounds awayClose, mid or far
MovementWhether, how fast and along what path a sound travelsStatic, slow or fast, plus path
EnvelopmentBeing inside the music rather than in front of it1 to 5
ClarityHow easily you follow each part1 to 5
MaskingOne sound hiding another, usually via shared frequencies and directionName the other stem

You reduce masking by separating parts in direction, height or distance.

Procedure

  1. Use headphones, because binaural rendering delivers 3D cues directly to each ear and room loudspeakers smear them. Record the device you use.
  2. Choose a quiet room. Background noise masks reverb tails and soft high frequencies that carry distance and height.
  3. Set a moderate level and keep it fixed. A louder version tends to sound better for a moment, which biases comparisons.
  4. Optionally enable a hearing compensation profile, and record whether you used it.
  5. Pass 1, whole mix: listen without soloing. Judge envelopment, the front, anything above you and where attention goes first. Write three sentences.
  6. Pass 2, solo layers: solo one stem or small group and record position, height, distance and width in one worksheet column per stem. Add a second stem and listen for masking.
  7. Pass 3, A/B: switch 3D positioning off and on, for example 20 seconds each, three times. Note what collapses and what opens up. Quick switching works because memory for small sonic details is short.
  8. Pass 4, format switch: play the same section in two spatial formats, such as a channel and object-based mix and Higher-Order Ambisonics, and compare clarity, focus, envelopment and masking.
  9. Pass 5, head tracking: turn your head slowly left and right and note whether sounds stay anchored and whether front and back become clearer.
  10. Stop, rest your ears in silence and complete your notes. In long sessions, take a short break every 20 to 30 minutes.

For each pass, draw a top-down map: front at the top, rings for close, mid and far, and a centre zone for sounds above. Mark a focused source with a dot, a wide source with an arc, a source above with a ring in the centre zone and movement with an arrow, and label each mark.

Next steps

Go deeper on the blog