INU — Support & User Manual

Stereo & surround → Dolby Atmos upmixer & spatial visualiser

Version 3.1.0 · Need a hand? support@demix.me

1. What is INU?

INU is a real-time upmixer that converts a stereo — or an existing 5.1 / 7.1 surround — source into a full surround or Dolby Atmos bed, from Quad and 5.1 up to 9.1.6, covering the bed, wide channels and height — inside your DAW. Two upmix engines share the same controls: NULLING, a lossless M/S encoding design (the “nulling” property) where the original stereo image is always recoverable from the output, and ENERGY, a power-complementary fold that keeps the front intact and wraps a decorrelated room around you — matched in loudness and timbre. A NULLING ⇄ ENERGY fader morphs between them. A real-time 3-D spatial energy visualiser lets you see exactly where energy is being placed in the room.

The output layout is determined automatically by what your DAW and track configuration reports. INU supports both flat surround — Quad/4.0, 5.0, 7.0, 5.1, 7.1 — and Dolby Atmos with height — 5.1.2, 5.1.4, 7.1.2, 7.1.4, 7.1.6, 9.1.6. On a layout without height the overhead sends fold into the surround bed; on a layout without a centre (Quad) or LFE (the .0 formats) those extractions fold into the L/R mains, and the controls for absent channels grey out automatically.

The two signal layers

  • Direct — the focused, centre-of-the-mix content. Push it outward to place it into surrounds and wides.
  • Diffuse — the width, reverb tails and ambient content. Push it outward for envelopment and space.

Centre and LFE are extracted from the Direct signal. The Null indicator confirms that the Direct/Diffuse split still recombines losslessly to your original stereo.

2. Installation & licensing

EditionFormats includedNotes
FreeStandalone application onlyRequires email registration. Periodic internet check.
PaidVST3 · AU · AAX · CLAP · LV2 · StandaloneUp to 3 machines. Periodic internet check.

Run the installer for your platform and follow the prompts. On Windows the installer is signed; a brand-new release may show a Microsoft SmartScreen “unrecognised app” notice until download reputation builds — this is normal for a newly-signed installer and does not indicate a problem.

Linux ships as a .tar.gz rather than an installer. Extract it, then copy the format(s) you use into your user plug-in folders and rescan in your DAW: VST3 → ~/.vst3/, LV2 → ~/.lv2/, CLAP → ~/.clap/. The Standalone runs directly (./INU) — no install needed. Reaper and Bitwig load VST3/CLAP; Ardour prefers LV2.

3. DAW compatibility & setup

INU accepts a stereo input — or an existing 5.1 / 7.1 surround bed to expand (a paid feature) — and automatically adapts its output to the channel layout your DAW and track report — any surround or Dolby Atmos layout from Quad/5.1 up to 9.1.6 (see §1 for the full list).

DAWFormat
Pro ToolsAAX (paid tier)
Logic Pro, Nuendo, Cubase, Reaper, Studio OneAU or VST3
Sequoia, SamplitudeVST3
Ableton Live, BitwigVST3 (automatic multi-bus mode)
Bitwig, Reaper, Studio OneCLAP also supported

Reaper

Layout negotiation requires the track/bus channel count to match the layout exactly. Set the track to the correct channel count before inserting INU:

  • 16 channels → 9.1.6
  • 12 channels → 7.1.4
  • 10 channels → 7.1.2
  • 8 channels → 5.1.2
  • 6 channels → 5.1
  • 4 channels → Quad / 4.0

Logic Pro

Logic has a few specific requirements — once they're met INU works normally:

  • Stereo track — insert INU on a stereo audio track, not a mono one. Logic filters the plug-in menu by a track's input format, and INU is a stereo upmixer, so on a mono track (Logic's frequent default) it won't appear. If you can't find INU, set the track's input format to Stereo on the channel strip.
  • Use an Atmos / height track — set Project Settings → Audio to a Dolby Atmos or height format (5.1.2, 5.1.4, 7.1.2, 7.1.4) and insert INU there. Logic offers plug-ins on its legacy surround tracks (Quad, 5.1, 7.1) selectively, by its own internal rules — independent of what a plug-in advertises — so INU (like many third-party plug-ins) may not appear on a legacy surround track even though it validates correctly. On an Atmos/height track it's fully available; on a no-height project the overhead sends fold into the surround bed. See Known issues for detail.
  • 7.1.4 is the ceiling — Logic doesn't support 9.1.6, so INU tops out at 7.1.4 (the wide and side-height channels aren't used).
  • Auditioning needs a signal — Logic only runs a track's plug-ins while audio is flowing, so INU's speaker test and the energy view need a signal playing. Park an audio clip on the track and press Play.

Requires INU build 109 or later. INU routes by channel type, so Logic's own speaker order (and the Default / ITU / WG-4 setting) is handled automatically.

Samplitude & Sequoia

In these hosts multichannel lives on buses and aux objects, not tracks — there is no surround track to create. Make a surround / Atmos bus or aux of a supported width (e.g. 7.1.4 or 9.1.6), insert INU on it, and feed it from a stereo track or send. INU reads the bus's L/R and renders the full field across every channel. If INU only outputs two channels, it's sitting on a stereo track — move it to the surround bus/aux.

Ableton Live & Bitwig (stereo multi-out)

Live and Bitwig present plug-in outputs as paired stereo buses. INU detects these hosts automatically and routes its 16 channels across 8 stereo buses (L/R, C/LFE, Lss/Rss, Lrs/Rrs, Lw/Rw, Tfl/Tfr, Tsl/Tsr, Trl/Trr) in its internal canonical order.

  • Bitwig exposes all eight buses at once — assign each to the correct speaker in your host.
  • Ableton Live enables an output pair only when you route it. Create an audio track, set Audio From → the INU track → the bus (e.g. Lss/Rss), and Monitor → In. Any pair you have not routed folds into the main L/R so nothing is lost; routing a pair out moves that content off L/R onto its own track (a move, not a copy — no duplication). Remove the receiver and it folds back into L/R.

If any channel comes out on the wrong speaker, use the Output routing matrix to correct it.

4. Interface at a glance

The editor is a fixed size and everything is colour-coded to the signal path:

ColourMeaning
RedDirect signal path — the focused (mono/centre) content spread into speakers
BlueDiffuse signal path — the stereo width/ambience spread as envelopment
YellowCentre channel extraction
PurpleLFE (subwoofer) channel extraction
GreyDecorrelation controls
GreenOSC headtracking active / Null check OK
OrangeEnergy engine (NULLING ⇄ ENERGY fold + Energy dials) · active quick preset · warnings

Layout: a top bar (headtracking left, INU logo centre, the tabbed OUTPUT panel right, quick presets either side), a main area of concentric arc sliders around the spatial sphere, the central spatial energy view, and a single-line info bar at the very bottom showing help for whatever control you hover. Click the INU logo to open the in-plug-in manual and the routing matrix.

5. How INU works — signal flow

  1. Stereo input is split into Direct (mono/centre-weighted) and Diffuse (stereo side/width) components.
  2. Optional Normalise continuously levels the rendered output toward the selected delivery target's integrated loudness (by driving the master Output Trim) — a monitoring aid, off for final render.
  3. LFE and Centre are extracted from the Direct signal first.
  4. The remaining Direct feeds the Direct layer; the Diffuse feeds the Diffuse layer.
  5. Each layer is distributed across the 9.1.6 canonical speaker grid using the Push parameters.
  6. Optional Decorrelation adds phase diversity between speakers for width and envelopment.
  7. In Binaural/Headtracking mode the entire 9.1.6 field is folded to stereo via VBAP.
The Null OK indicator confirms the encode-decode path is lossless. It goes red when LFE Push is used in Additive mode, or Decorrelation is above zero, because those intentionally add energy or phase that is not part of the reversible stereo transform — this is expected behaviour, not an error. Centre Push and Subtractive-mode LFE keep it green.

6. Controls reference

Direct layer

Carries the focused, centre-of-the-mix content. Increasing push spreads that focused energy outward.

ControlRangeDefaultWhat it does
Direct Push0 – 10.35How much Direct energy is pushed from the front L/R bed into surrounds and wide channels.
Direct Height Push0 – 10.15Of the pushed Direct energy, how much rises into the overhead height speakers.
Direct Height HPFOff – 20 kHzOffHigh-pass before the height send. Keeps low end anchored in the bed.
Direct Front/Rear−1 → +10Bias Direct energy toward front (negative) or rear (positive). On a layout with a single surround pair (≤5.1) the positive half instead wraps the front image into the surround for a fully enveloping feel, so the whole travel stays useful.
Direct FR HPFOff – 20 kHzOffHigh-pass before the front/rear send.

Diffuse layer

Carries the width, reverb tails and ambient content — the sense of envelopment and space.

ControlRangeDefaultWhat it does
Diffuse Push0 – 10.85How much Diffuse energy is spread into surrounds, wides and height. Higher = more envelopment.
Diffuse Height Push0 – 10.20Of the pushed Diffuse energy, how much goes to height channels.
Diffuse Height HPFOff – 20 kHzOffHigh-pass before the diffuse height send.
Diffuse Front/Rear−1 → +10.10Bias Diffuse energy toward front (negative) or rear (positive). As with Direct, on ≤5.1 the positive half wraps the image into the surround (enveloping) instead of the absent rear.
Diffuse FR HPFOff – 20 kHzOffHigh-pass before the diffuse front/rear send.
Higher Diffuse Push with active Decorrelation gives the most convincing sense of room envelopment. Start around 0.7–0.9 and adjust to taste.

Energy engine — the NULLING ⇄ ENERGY fold

The large NULLING ⇄ ENERGY arc dial wrapping the spatial view morphs between INU's two upmix engines. NULLING (left, default) is the lossless coherent engine described above — the Direct/Diffuse split recombines exactly to your stereo. ENERGY (right) is a power-complementary “energy null”: it keeps the front image intact and sends a decorrelated copy of the room around you, matched in loudness and timbre rather than sample-for-sample, so it stays mono- and downmix-safe while sounding fuller and more enveloping. Drag the dial to blend the two; the inactive engine's dials dim and desaturate to greyscale. On a fresh insert the Energy voicing (Source 50 %, HP Send 110 Hz) makes a NULLING → ENERGY sweep a tonal match with clearly more envelopment.

ControlRangeDefaultWhat it does
Energy Room Push0 – 100%35%How much signal moves into the decorrelated room vs staying up front. Power-complementary, so loudness is preserved as you push.
Energy Source Push0 – 100%50%The room feed: left = whole mix (fullest), right = side-only (keeps the centre locked to the front).
HP Send Crossover0 Hz – 20 kHz110 HzBass below this stays up front; only higher frequencies go to the room. Fully left (0) = full-band send.
Energy Front/Rear Push−100% → +100%−6%Room placement: left = wide/front, centre = sides, right = rear on 7.1+ — or fully enveloping on ≤5.1 (with a single surround pair the positive half wraps the front image around you instead of the absent rear).
Energy Height Push0 – 100%10%Lifts the room into the height layer (needs 7.1.4 / 9.1.6).
Room Decorrelation0 – 100%100%Decorrelation amount on the energy room send. Diffuse-nature, so it stays safe in binaural.
Front Decorrelation0 – 100%0%Optional front-image width via an L/R all-pass (bass stays mono). Blurs the front by design; auto-bypassed in headtracking.

Centre channel

Extracts a band-passed portion of the Direct signal into the dedicated centre speaker. Useful for anchoring dialogue or a lead vocal.

ControlRangeDefaultWhat it does
Centre Push0 – 10Amount of filtered Direct sent to the centre speaker. Subtractive: removed from the bed.
Centre BP Filter20 Hz – 20 kHz1 kHzCentre frequency of the band-pass extraction filter.
Centre Q0.05 – 100.71Bandwidth of the band-pass. Higher Q = narrower band.

LFE / subwoofer

Extracts low-frequency content from the Direct signal into the LFE channel. The small + checkbox in the centre of the LFE arc toggles the extraction mode.

ControlRangeDefaultWhat it does
LFE Push0 – 10Amount of low-passed Direct sent to the LFE channel.
LP Filter5 – 130 Hz120 HzCutoff of the low-pass extraction filter.
LFE Q0.05 – 100.71Q of the low-pass filter.
LFE Mode ☑toggleSubtractiveSubtractive (default): LFE removed from the bed. Additive: LFE is an extra send; bed unchanged.
In Additive mode the Null indicator goes red when LFE Push > 0 — the extra LFE send adds energy outside the reversible transform. Subtractive mode keeps nulling (the low end is losslessly moved out of the bed, not added), so it stays green.

Decorrelation

Decorrelation adds frequency-dependent phase differences between speaker channels using multi-band all-pass networks. This creates perceptual width and envelopment without affecting spectral content — you cannot hear it on a single channel, only as a spatial impression across the field.

Important: Direct-layer decorrelation is automatically bypassed when Binaural/Headtracking mode is active — both the Direct Decorrelation dial and the Energy Front Decorrelation dial — because the VBAP fold causes phase-shifted front channels to comb-filter as the head rotates (they grey out while headtracking is on). The diffuse-nature decorrelators — Diffuse Decorrelation and the Energy Room Decorrelation — stay live and are always safe to use.
ControlRangeDefaultWhat it does
Direct Decorrelation0 – 10Global wet amount on the Direct layer. 0 = fully dry.
Diffuse Decorrelation0 – 10Global wet amount on the Diffuse layer. Safe in binaural mode.
HP Filter20 Hz – 20 kHz300 HzDecorrelation only applied above this frequency. Preserves low-end mono compatibility.
Wide Trim0 – 20.65Scales decorrelation on the wide channels (Lw/Rw). Reduce if wides feel too phasey.
Front/Rear−1 → +10Bias decorrelation toward front (negative) or rear (positive).
Height Bias−1 → +10Positive = more decorrelation on height; negative = more on the bed.

Output trims, Link & the Downmix reference

Two small trim sliders on the OUTPUT panel's Fold tab fine-tune the level and balance of the multichannel upmix — most useful for an A/B level-match against the source, which can read a touch quieter as energy spreads into the surrounds.

ControlRangeDefaultWhat it does
Direct trim−12 → +12 dB0Scales the whole Direct layer at the output.
Diffuse trim−12 → +12 dB0Scales the whole Diffuse layer at the output.
Link ☑toggleOnOn = both trims move together as one master level-match. Off = trim the Direct/Diffuse balance independently.

Downmix reference

The Downmix button (right of the Direct row) shows the current reference and a caret; click it to expand the Loudness panel down and reveal four references plus the fine trims. A reference pre-boosts the surround-plane channels in the multichannel output (the inverse of the attenuation a standard downstream downmix applies), so a naive downmix folds them back to unity and stays level-matched to your mix. This is a delivery / reference choice — it does raise those channels in what you monitor (surrounds and heights read louder). Off = unity (bit-identical). The boost only ever hits what a downmix attenuates: side/rear surrounds + top-side/top-rear heights. Centre, front L/R, wides and top-front fold to the front, so they always stay at unity.

ReferenceSurroundHeightWhat it is
Off00No compensation (unity).
BS.775+3 dB+3 dBITU-R BS.775, the faithful / downmix-safe default (the inverse of the −3 dB a standard downmix applies). BS.775 is a 5.1-era spec and doesn't define heights; the height figure is our reasonable extension.
Atmos+6 dB (5.1)
+5 dB (7.1.x+)
— (5.1)
+3 dB (7.1.x+)
Home / equal-level reference matched to the Dolby Atmos auto-downmix balance.
Cinematic+6 dB— (5.1)
+3 dB (7.1.x+)
A deliberately surround-forward feel for the film crowd (Penteo/Halo-style). A taste, honestly labelled — not a standard.

Why Atmos differs by layout — the fold cascade. An Atmos renderer doesn't drop straight to stereo; it collapses in stages — 9.1.6 / 7.1.4 → 7.1 → 5.1 → 2.0, with the tops folding into the bed first. Each surround-collapsing stage costs a channel about −3 dB, so a channel passing through more folds needs more make-up: a 5.1 surround sees one fold (+3 dB, = BS.775), a 7.1 surround sees two (≈ +6 dB), and heights cascade through an extra fold again. The Atmos figures start from that math but are trimmed to match the balance a real Atmos auto-downmix produces rather than pure unity survival (Dolby folds the tops gently) — treat them as sensible starting points and adjust to taste.

The figures are layout-aware — INU reads your current output layout and picks the bed vs immersive values. The reference simply sets the two fine trims below it; nudge either afterwards and the button reads Custom.

TrimRangeWhat it does
Surround−6 → +6 dBBoost on the side/rear surround (and top-side/top-rear) pairs. Overrides the reference value.
Height−6 → +6 dBBoost on the height/top pairs. Greys on flat layouts (5.1 / 7.1) with no height channels.
Your adjustments are remembered. The reference and its two trims are ordinary parameters — they save with your project/session and are captured in every P1–P8 preset, so a hand-tuned Surround/Height comes back exactly as you left it (the button reads Custom when the values match no built-in reference). Nothing extra is stored — the reference label is re-derived from the saved trim values on recall, so older projects and presets load unchanged.
The trims and downmix reference shape the multichannel render only — they have no effect in binaural/headtracking mode and grey out there. They also never move the Null lamp: it always measures the pre-trim signal, so monitoring stays honest.

6b. Upmixing a 5.1 / 7.1 source (paid)

As well as upmixing stereo, paid INU can take an existing 5.1 or 7.1 bed and expand it to immersive (up to 9.1.6). The original bed passes through untouched — INU generates only the channels the target layout has but the bed does not (the added heights, side/back surrounds and wides).

  • Set it up. Put INU on a track whose width is your target layout (e.g. 7.1.4 / 9.1.6) carrying the surround source, open Settings and set Upmix Input Format to 5.1 or 7.1 (leave it Auto / Full for a stereo source — the classic stereo upmix). The format strip reads e.g. 5.1 > 9.1.6 while upmixing.
  • Faithful → Immersive master. In this mode the large arc dial (labelled FAITHFUL / IMMERSIVE) becomes a single master over the generated field: hard-left = the bed as-is (nothing added), hard-right = the full generated immersive field. It opens at a healthy amount so you hear the expansion straight away; pull it toward FAITHFUL for a lighter touch.
  • Shape it with the Energy dials. Energy Room (surround amount), Energy Height, Energy Front/Rear (placement), Energy Source (whole-mix → side/ambient), the HP Send Crossover (keeps bass in the bed) and Room Decorrelation (width) fine-tune within the master amount. The stereo-only Nulling dials (Direct/Diffuse Push and their decorrelation) grey out here — hover any greyed dial and the info bar tells you why.
  • Centre / LFE trims. The Centre Push and LFE Push dials become bipolar ±dB level trims on the passed-through bed centre / LFE (their derivation filters grey out — a discrete bed has nothing to derive).
  • Downmix reference still applies. The Fold tab's Downmix reference (Off / BS.775 / Atmos / Cinematic + Surround/Height trims) pre-compensates the immersive output for a downstream fold on a bed input too — pick a reference to target a delivery downmix, or leave it Off to keep the bed bit-verbatim.
  • Non-standard channel order? Use the Input Routing (upmix bed) matrix in Settings to point each bed channel at the right input.
Surround-input upmix is a paid feature; the free standalone folds the front L/R only. It isn't available in Bitwig / Ableton, which feed plug-ins stereo.

7. Working with the controls

  • Drag any arc to change its value (standard/coarse).
  • Ctrl-drag (Cmd-drag on macOS) for fine adjustment.
  • Double-click a parameter to reset it to its default value.
  • Double-click the spatial sphere to reset the view (orientation and zoom) — it does not touch your parameters.
  • The info bar at the bottom shows a description of whatever control you hover, tinted by its signal-path colour.
  • Controls that are inactive in the current mode both dim and desaturate to greyscale — a disengaged upmix engine's dials, the Direct-layer decorrelation while headtracking, or a top-bar slider that doesn't apply — so the live controls read at a glance.

8. The spatial energy view

The central sphere is a real-time OpenGL dot-cloud showing signal energy distributed across the speaker field. It updates continuously from the audio thread.

  • Red dots — Direct energy, focused at speaker positions.
  • Blue dots — Diffuse energy, spread broadly as ambience.
  • Dot brightness and size scale with signal level. Panning the stereo input moves energy left/right across both clouds.
  • The yellow ring marker is the energy centroid — the moving centre-of-mass of all energy in the field. It drifts toward wherever the mix is loudest, giving the overall spatial balance at a glance.
  • A faint square floor grid beneath the sphere is a fixed ground reference, so the 3-D orientation reads clearly as you rotate.

Interacting: drag to rotate, scroll to zoom, middle-click (scroll-wheel click) or Cmd/Ctrl+click to pop the sphere out into a resizable floating window; the same gesture inside the pop-out closes it. Click a speaker to send a test tone to that output (click it again to stop). During silence, two small birds glide idly inside the sphere and fade out when audio resumes.

Some hosts — Samplitude, Sequoia, and Nuendo on an idle bus among them — only run plug-ins while the transport is rolling, so if a speaker audition is silent, press Play (or enable input monitoring). The info bar at the bottom of the plug-in reminds you when this is the case. The energy view and the halo-intensity slider likewise need audio flowing to react.

Performance & quality (Windows laptops): the spatial view is GPU-accelerated. On a dual-GPU laptop, Windows often runs your DAW — and INU's view — on the power-saving integrated GPU, which can feel heavy at High quality. To assign the discrete GPU: Settings → System → Display → Graphics → Add desktop app, pick your DAW's .exe, then Options → High performance (the NVIDIA/AMD GPU), and restart the DAW. The pop-out shows a small fps readout (top-right) so you can gauge each quality setting; drop to Medium/Low if needed. It all runs off the audio thread, so it never affects audio — only UI smoothness.

Halo intensity, response & quality: two small sliders sit in the strip below the view, with the layout format label (e.g. 9.1.6) between them. Both are purely visual — they do not affect audio — and both are saved with the session and every preset. Intensity (left) trims how bright the energy halo appears; brightness is held consistent regardless of window size or input gain, so the halo looks the same in the inline view and the large pop-out. Response (right) sets how fast the halos react to the audio: right = snappier, left = smoother. Spatial view quality (Low / Medium / High), set in the Settings tab of the in-plug-in manual (click the INU logo to open it), controls dot-cloud density — higher is smoother and denser but asks a little more of the GPU; Medium suits most systems. The Settings tab also holds the delivery-target selector, the “Show dial values as %” toggle, the “Even intensity when Energy spreads” option and the “True-Peak trim to the target ceiling” toggle; a Diagnostics tab (dev/beta builds) below it shows a live host/layout/routing report with a “Send debug report” button. (The downmix reference and its Surround/Height trims live on the OUTPUT panel's Fold tab, not in Settings.)

9. Headtracking & binaural monitoring

INU can fold the full 9.1.6 field to binaural stereo in real time, enabling headphone monitoring with spatial awareness. Two modes are available.

OSC headtracking

Enable with OSC HT. INU listens on a UDP port for OSC rotation messages and folds the field to binaural in real time. The default port is 8000, changeable in the text field — the receiver restarts automatically. The OSC status lamp shows: grey = inactive, red = listening but no data, green = receiving within the last 2 seconds. Crossfeed adds inter-ear crosstalk for a more natural headphone image.

INU uses the Supperware Bridgehead head-tracker profile:

  • Address: /ypr
  • Arguments: three float32 values — yaw, pitch, roll
  • Units: degrees (converted to radians internally)
  • Transport: UDP, default port 8000

Bridgehead profile entry (add to Bridgehead’s Profiles.txt):

INU    /ypr    yaw,pitch,roll    local 8000

Any head-tracker app that can send /ypr with three float32 degree values over UDP is compatible.

Emulate headtracking

Enable with Emulate HT, then drag the spatial view sphere to simulate head rotation — the binaural fold updates in real time. Useful for auditioning on headphones without a physical tracker.

Head View & moving within the sphere

Toggle Head View (alongside OSC HT or Emulate HT) to switch the sphere’s drag convention so it represents your head rotating, not the world. With Head View active and a headtracking mode engaged, the cursor arrow keys move your listening position within the sphere: Up/Down = forward/back, Left/Right = left/right. Combined with head rotation (yaw and pitch by dragging in Emulate HT, or full yaw/pitch/roll from a hardware tracker in OSC HT) you can audition the fold as you turn your head and shift position. This is positional movement on the horizontal plane only (no up/down), so it is not full six-degrees-of-freedom (6DOF) tracking.

10. Quick presets

Eight slots (P1–P8) sit either side of the sphere, four per side. Left-click a filled slot to recall it, double-click to save the current state into it, right-click to clear it. A preset captures everything you can set: all parameters (including headtracking mode, port, crossfeed and Head View, the Direct/Diffuse output trims, Link and the downmix reference), your output routing (matrix overrides and the C+LFE collapse) and the spatial-view density, intensity and response trims.

Presets are kept in a personal file so they follow you between projects — %APPDATA%\INU\quickpresets.xml on Windows, ~/Library/Application Support/INU/quickpresets.xml on macOS, ~/.config/INU/quickpresets.xml on Linux.

  • In a new, unsaved project your eight slots load from this file, so P1–P8 are ready the moment you add INU. Saving or clearing a slot rewrites the file straight away, and every new instance you open from then on sees the change.
  • In a saved and re-opened project a snapshot of all eight slots is also stored inside the project, so your presets travel with the session and come back exactly as you left them — even on a machine that has never seen your personal file. The trade-off: a project keeps the preset versions it was saved with, so if you later change a slot in a new session, older projects still recall the version they were saved with rather than your newer one.

11. Routing matrices

You normally never need this. INU routes every recognised host automatically, so the default (all-green “auto”) mapping is already correct. The matrix is an escape hatch for the rare case where a channel comes out on the wrong speaker — there is no universal standard for Atmos channel ordering, and JUCE, DAWs and plug-ins can each order a layout differently. Open it from the Settings tab of the in-plug-in manual (click the logo → Settings). If anything ever looks wrong, “Reset to auto” restores the default, and your overrides are saved with the project.

  • Rows are INU’s internal channels present in the current layout; columns are the host’s physical outputs (auto-sizing to the channel count). Both axes are labelled with your DAW’s own speaker names — for hosts that only number their channels, INU’s canonical names are shown instead.
  • Green dot — the channel’s automatic output (no override).
  • Left-click a cell to override that channel’s output → the dot turns blue.
  • Left-click a channel’s own green (auto) cell to revert it.
  • Orange — a routing conflict (two channels aimed at the same output).
  • Right-click a channel row to solo its output (its dot turns yellow) — every other output mutes so you can confirm what physically reaches it. Right-click again (or close the tab) to clear.
  • Reset to auto clears all overrides. Canonical 1:1 — shown only on numbered/position-based hosts (e.g. Reaper, Bitwig) — forces INU’s canonical order (channel i → output i). It is hidden on type-aware hosts (Steinberg, Pro Tools, Logic, MAGIX), which route by channel type, where an identity override would only misroute.

Overrides are saved with the session/preset. The row and column labels are your host’s own speaker names, so you can match them directly against your DAW’s output meter.

Unrouted buses on Ableton Live

Ableton enables an output pair only when you route it to a track, whereas Bitwig exposes all eight at once. INU handles this automatically: any bus you have not routed folds into the main L/R, so the plug-in's own track plays a complete downmix rather than dropping channels. Route a pair out to its own track and that content moves off L/R onto the discrete track — a move, not a copy, so it never plays in two places. Removing the receiver folds it back into L/R. No matrix changes are needed for this.

Collapse C & LFE to dual-mono (stereo hosts only)

On stereo multi-out hosts (Bitwig, Ableton) a checkbox appears at the bottom of the matrix. When enabled, the C and LFE outputs both carry (C + LFE) × 0.707 (−3 dB), so a naive stereo fold of the multi-out keeps the centre and LFE content centred at correct level rather than hard-left/hard-right. Default off; the discrete C and LFE remain on their own channels unless you enable it.

Input routing (surround-input upmix bed)

A second matrix appears while a surround-input upmix is active (Upmix Input Format set to 5.1 / 7.1 — see §6); otherwise it shows a short “no upmix active” note. Rows are INU's bed channels; columns are your host's input channels. INU reads a standard 5.1 / 7.1 source automatically — use it only to correct a source whose channels are in a non-standard order (for example an SMPTE 5.1 with Ls/Rs on channels 3–4). Reset to auto restores the default.

12. The OUTPUT panel — loudness & QC

IndicatorMeaning
The OUTPUT panel (top right) is tabbed — Loud / QC / Fold. Two status dots stay visible on every tab: an Upmix QC dot (worst of coherence / phase / null) and the Safe-to-print dot.
Delivery targetA dropdown on the Loud tab (mirrored in Settings) sets the standard the output is graded against — an integrated-loudness target and a True-Peak ceiling: Streaming -14, Apple Music -16, Atmos music -18, EBU R128 -23, ATSC A/85 -24, Netflix film -27, or Off/Custom. It drives the readout colours, the Safe lamp, the Normalise leveller and the True-Peak-trim ceiling.
Loudness readoutFull BS.1770 on the delivered output (at your actual layout — surround/height +1.5 dB, LFE excluded): Integrated and True-Peak (max held, live in brackets) as the large compliance values, colour-graded against the target (green in tolerance, amber close, red/blue over/under), with Short-term, Momentary and LRA beneath. R resets the integrated / LRA / TP-max measurement.
NormaliseA continuous monitoring leveller: while engaged it drives the master Output Trim to bring the rendered output to the selected target's integrated loudness (−14 / −18 / −23 …), so you can A/B at delivery level — watch it move on the Output Trim slider. Commit it with the T (Apply to output trim) button on the Loud tab; otherwise off for final render — see the note below.
● Safe-to-printGreen when the output meets the target (integrated in tolerance and True-Peak under the ceiling) with no dynamic gain, so an offline render complies as-is. DYN = met only via the Normalise leveller (dynamic); OVER/CHECK = not yet compliant.
QC tab: CoherencePer-band coherence of the output, low to high: red = correlated (mono / combing risk), green = decorrelated. Folding toward Energy greens it.
QC tab: ● φ (Surround phase)Worst inter-channel correlation across the surround/height/wide pairs. Green = mono-safe · amber = decorrelated & mono-acceptable · red = anti-phase (correlation → −1) and will cancel on a mono fold (QC risk). Raise Diffuse Decorrelation to climb toward green. Grey in binaural. See below.
QC tab: ● Null?Green = the Direct/Diffuse split recombines losslessly to the original stereo. Red (expected) = LFE Push in Additive mode, or Decorrelation, is above zero. Centre Push and Subtractive-mode LFE stay green. If it goes red with all controls at zero, please report it.
Fold tab: trims · Link · DownmixOutput level-match (Direct / Diffuse trims + Link) and the Downmix reference — expand it for Off / BS.775 / Atmos / Cinematic + Surround/Height trims (see §6). Multichannel render only (greyed in binaural).

Normalise is a live monitoring leveller, not a render-time loudness pass. It continuously drives the Output Trim from a live loudness measurement — so leave it off for your final render and set delivery level with a proper loudness / mastering pass (or the output trims). To bake a compliant static level for print, press T (Apply to output trim) on the OUTPUT panel's Loud tab — the render-safe bridge: it commits the leveller's converged gain (or the suggested trim toward the target) into the static Output Trim and switches the leveller off. It may take a few seconds to settle on very short loop points.

True-Peak trim

A Settings option — “True-Peak trim to the target ceiling” — transparently shaves stray inter-sample peaks on the delivered output down to the ceiling of the loudness target you've chosen (e.g. −1 dBTP), so you can keep your loudness while meeting a True-Peak spec without pulling the fader or slamming a limiter. It only acts on actual peaks — the rest of the signal passes through untouched — and the Trim readout shows the gain reduction it's applying (max held, live in brackets). Unlike Normalise this is a static process (a fixed per-sample non-linearity), so it is safe to print on a final render.

The True-Peak trim is built on the ClipOnly algorithm by Chris Johnson / Airwindows, used under the MIT licence. Our thanks to Chris for making his work openly available — airwindows.com.

Surround phase & mono-compatibility (the φ lamp)

INU's diffuse layer is the side signal (S = ½(L−R)). To stay losslessly reversible (the null), every pair is encoded as L = direct + S, R = direct − S — so a pair carrying the diffuse signal is anti-phase between its L and R by design. A surround pair that receives diffuse with little accompanying direct reads negatively correlated, and pure ±S reads −1 (it cancels on a mono fold — a QC red flag for broadcast/streaming delivery).

The φ lamp shows the worst such correlation in real time, calibrated to actual mono risk: green (≳ −0.3) = mono-safe; amber (−0.3…−0.6) = decorrelated and mono-acceptable (only a ~1.5 dB partial loss — the normal working zone for a wide upmix); red (≲ −0.85) = near-cancellation, little or no decorrelation on strongly diffuse surrounds (the QC danger to avoid before delivery).

The main lever is Diffuse Decorrelation: it applies an independent all-pass network per channel, so the L and R of each pair stop being ±copies and become mutually decorrelated — correlation climbs red → amber → green as you raise it. (INU decorrelates the low band only ~45% by design to keep the bass clean, so a pure-diffuse pair settles in amber rather than full green — that's expected and mono-acceptable.) Letting some direct back into the surrounds (ease Front/Rear off the extremes, or raise Direct Push) correlates positively and lifts the lamp further; the decorrelation HP filter sets how low the decorrelation reaches. The trade-off is that decorrelation breaks the exact null (the Null lamp goes red); for a delivery mix, mono-compatibility is the priority, so decorrelation-on is the correct posture for post-production. INU ships null-exact by default (Null green, decorrelation off), with Diffuse Front/Rear at 0 so a fresh insert never pushes diffuse into the empty rear surrounds — the φ lamp is your cue to add Diffuse Decorrelation before delivery.

13. Getting started — a suggested first session

  1. Insert INU on a stereo track and configure your DAW output bus to a surround or Atmos layout (anything from Quad/5.1 up to 9.1.6). (Reaper: set the track channel count first — see §3.)
  2. Leave all controls at defaults. Play audio and watch the spatial view fill with red (Direct) and blue (Diffuse) energy.
  3. Raise Diffuse Decorrelation to around 0.4–0.6 and listen to the increased envelopment. Keep the decorrelation HP Filter at 300 Hz to preserve bass mono compatibility.
  4. Try Normalise if your source is at broadcast levels.
  5. Add LFE Push at 0.2–0.5 for music with significant bass. Use Subtractive mode (default) to avoid bass build-up.
  6. Save to P1 — double-click the P1 button.
  7. With headphones, enable Emulate HT and drag the sphere to hear the binaural fold.

14. Troubleshooting & FAQ

INU won’t load / no multichannel output. Confirm your track/bus is a supported surround or Atmos layout — anything from Quad/5.1 up to 9.1.6 (see §1). In Reaper, set the track channel count to match the layout exactly before inserting INU.

In Ableton Live I route an output pair to a track but hear nothing. Live enables an output pair only when it is routed. On the receiver track set Audio From → the INU track → the bus, set Monitor → In, and play spatial source material. Unrouted pairs fold into the main L/R, so route the ones you want discrete (see §11).

Only two channels out in Samplitude, Sequoia, or Logic. These hosts don’t widen a stereo track through a plug-in. Insert INU on a surround / Atmos bus or aux (Samplitude and Sequoia have no surround tracks) or an Atmos/surround track (Logic), fed from a stereo source — see DAW setup. Logic also needs build 109+ for INU to appear on surround tracks.

INU won't insert on an Aux in Pro Tools. In Pro Tools (AAX) INU is a stereo-in plug-in — it upmixes a stereo source and widens the output to your surround/Atmos path. So it can only be inserted on a stereo track or Aux; a surround- or Atmos-width Aux won't offer it. To upmix group submixes, send them to a stereo bus, feed a stereo Aux Input from that bus, then insert INU choosing the output format you want from the list — the entries read INU (stereo/7.1.4), (stereo/9.1.6), etc., i.e. stereo-in to that format out. Finally set that Aux's output to your surround/Atmos bed. So you never create the Aux as surround — you create it stereo and pick the target format when loading INU. (Pro Tools is stricter here than Logic, which does place INU on surround tracks — see the note above.)

Clicking a speaker to audition is silent, or only works once. Some hosts only process plug-ins during playback, so a test tone can’t sound while stopped. Press Play (or enable input monitoring), then click the speaker — the plug-in’s info bar tells you when this is the reason.

A speaker plays the wrong channel (e.g. Bitwig/Ableton). Open the Routing tab and remap. Right-click a channel to solo its output and confirm what reaches each speaker, then left-click cells to correct the mapping.

The Null indicator is red. Expected whenever LFE Push is used in Additive mode, or Decorrelation is above zero. Centre Push and Subtractive-mode LFE do not affect it. Only report it if it’s red with every control at zero.

Direct decorrelation seems to do nothing in headtracking mode. It’s automatically bypassed in binaural mode to avoid comb-filtering during head rotation. Use Diffuse decorrelation, which stays active.

The stereo fold sounds lop-sided on a stereo multi-out host. Enable Collapse C & LFE to dual-mono in the routing matrix (§11).

The spatial view feels heavy or the sphere is jerky to drag (Windows laptop). On dual-GPU laptops Windows may run your DAW on the integrated GPU. Assign the discrete GPU via Settings → System → Display → Graphics (see §8), and/or lower the spatial-view quality to Medium/Low in the Settings tab. The pop-out's fps readout shows the effect.

SmartScreen warns about the installer on Windows. Normal for a newly-signed installer until download reputation accrues; it is not a malware warning.

Licence prompts after being offline a long time. Both tiers require a periodic online check. Reconnect to re-validate.

15. Support & feedback

When reporting a bug, please include:

  • OS and version (macOS 13+, Windows 10/11, or your Linux distro)
  • DAW name and version
  • Plug-in format used (VST3, AU, AAX, CLAP, LV2, Standalone)
  • A brief description of what you did, what you expected, and what happened
  • If the issue involves routing, note the output bus configuration in your DAW (a screenshot of the Routing matrix helps)

Contact support@demix.me.

16. Reference tables

Canonical channel order

INU’s internal 9.1.6 channel order:

#Channel#Channel
1L (Left)9Lw (Wide Left)
2R (Right)10Rw (Wide Right)
3C (Centre)11Tfl (Top Front Left)
4LFE12Tfr (Top Front Right)
5Lss (Left Surround Side)13Tsl (Top Side Left)
6Rss (Right Surround Side)14Tsr (Top Side Right)
7Lrs (Left Surround Rear)15Trl (Top Rear Left)
8Rrs (Right Surround Rear)16Trr (Top Rear Right)

Smaller layouts use the relevant subset (e.g. 7.1.4 omits the wide channels and the side-height pair). On surround (2D) layouts the height channels fold into the surround bed; Quad has no centre and the .0 formats have no LFE, so those extractions fold into the L/R mains.

17. Known issues & changelog

Current known issues and recent fixes for INU, updated as new versions are released. For what's planned next, see the INU roadmap.

Loading…