Twenty-odd small tools
that stop the day going sideways.

Reads Broadcast Wave files with the full CHNA and AXML description: every object, its position over time, its gain, and the block structure. Objects arrive as objects, not as channels with a note attached.

Export writes a delivery master that other immersive tools open without complaint. A file that goes in and comes back out is identical — that is tested on every build, not assumed.

Channels

Up to 128, RF64 for files past 4 GB

Metadata

chna and axml, full block-format parsing

Beds

Preserved as beds, not exploded into objects

Round-trip

Lossless — verified bit for bit

IAMF is the royalty-free immersive container from the Alliance for Open Media. Eclipsa Audio is the consumer name the same technology carries on some televisions and platforms — supporting one supports the other, with no extra work.

Export is bit-exact against the reference encoder, and can mux straight into MP4 for upload.

Encoder

Reference implementation, bit-exact output

Container

Raw IAMF or muxed into MP4

Layouts

Channel-based and object-based mixes

Direct path

ADM master in, IAMF out, without a detour

Immersive Audio Bitstream, the cinema format. Frame-locked to the picture rate, written as a flat bitstream ready for packaging.

Standard

SMPTE ST 2098-2

Frame rates

23.976 to 60, matched to picture

Direction

Import and export

Renders the scene to the Auro-3D speaker layouts, which put a height layer directly above the main layer rather than spreading it overhead. Written as plain multichannel audio in the documented channel order.

Layouts

9.1, 10.1 and 11.1

Output

Multichannel audio in the layout's channel order

Note

These are speaker layouts. Nothing here is a compatibility claim about any encoder or decoder

Renders the scene to the full 22.2 speaker arrangement used in Japanese broadcast — three height layers, twenty-four channels, in the documented channel order.

Channels

24, in NHK order

Layers

Top 9, middle 10, bottom 3, plus 2 LFE

Drop a 12-channel WAV and get twelve placed objects, named by their channel role, sitting where that layout says they belong. Ambisonics is decoded to a virtual speaker array first, so the scene stays editable rather than arriving as a locked field.

Layouts

Stereo to 9.1.6, plus custom orders

Ambisonics

First to third order, ACN/SN3D

Export

Back to any of the same layouts

SMX keeps everything editable and uncompressed — this is what you archive. SMXD is the delivery: audio encoded, objects split into independently addressable blobs, everything baked that needs baking.

Container

Plain ZIP — openable with any unzip tool

Codec

Opus or PCM, chosen at export

Streaming

Per object, so silence costs nothing

Licence

Free to read, write and ship

Renders the object scene to a fixed layout using the same panning path the players use, so the fold matches what a listener actually hears rather than approximating it.

Height layers are folded with a proper filter rather than summed flat, and bass management is applied where the target layout expects it.

Targets

1.0 mono, 2.0, 5.1, 5.1.2, 5.1.4, 7.1, 7.1.2, 7.1.4, 9.1.6

Loudness

Measured per fold to BS.1770-4

Bass

Optional management with adjustable crossover

Distance

Attenuation per doubling, adjustable, 6 dB is the physical value

Folds every object to two ears through the same path the mobile player uses, so what you check is what a listener gets. Useful as a deliverable and as a sanity test.

Profiles

Several head models, switchable

Reference

Loudness measured against the stereo fold, not the binaural one

Reads a master and judges it against the requirements of each destination: sample rate, bit depth, integrated loudness with tolerance, true peak ceiling and loudness range. You get a pass, a warning or a fail per target, with the number that caused it.

It does not fix anything. A tool that quietly changes your master is a tool you cannot trust.

Targets

Streaming, Apple, Atmos, EBU R128, CD, Club

Measures

Integrated LUFS, LRA, true peak, sample rate, bit depth

Standard

ITU-R BS.1770-4

Output

Per-target verdict with the failing value named

Encodes your file with the actual codec and bitrate a platform uses, decodes it again, and compares. You get the peak that appeared after encoding — which is usually higher than the original — and the level of the error signal, plus A/B playback between the two.

The point is the true peak. A master that sits at −0.1 dBTP frequently clips after lossy encoding, and that is where distortion complaints come from.

Presets

12, covering Apple Music, Spotify at four rates, YouTube Music and Video, Tidal, plus a worst case

Codecs

AAC, Vorbis, Opus at the platform's real bitrate

Reports

Peak and true peak before and after, error signal level

Listening

A/B switch between original and coded

Splits the file into bands, measures the energy in each, and checks phase correlation where it matters. A bass line that cancels in mono is inaudible on a club system with mono subs — this finds it on your desk instead of at the gig.

Bands

Sub, bass, low mid, mid, high — energy per band

Phase

Correlation per band, mono compatibility flagged

Verdict

Named problem, not a score out of ten

Aligns two files, inverts one, sums them, and reports what is left. Silence means identical. Anything else is the difference, and the level tells you whether it is a rounding error or a real change.

What it is for: proving that a round-trip through another tool changed nothing, or finding out exactly what a plugin did.

Alignment

Automatic offset detection by cross-correlation

Reports

Residual level in dB, plus where the difference sits in time

Measures inter-sample peaks the way BS.1770 says to, rather than reading sample values and hoping. Peak-safe applies a ceiling on export without touching anything below it.

Method

4× oversampled, BS.1770-4 kernel

On export

Optional ceiling, applied per file

Note

A positive dBFS peak on a summed multichannel file is normal, not a fault

Folds the scene to every smaller layout and compares loudness and energy distribution against the full render. Shows you what a phone, a soundbar or a stereo TV will actually do with your mix.

Compares

Full render against each fold, side by side

Flags

Elements that vanish or dominate after folding

Load a commercial reference and switch between it and your mix at matched loudness, so you are comparing balance rather than level. Reference Plus goes further: it profiles both, segments them, matches comparable sections, and writes out where yours differs and by how much.

Matching

Loudness-aligned switching, no level bias

Profile

Spectral balance, dynamics and stereo width per section

Output

Written notes and suggestions, not a magic button

Scans a file for cut points and writes out the pieces, each with a name built from the source and a timestamp or an index. Handles a live recording, a stem export or a sample pack the same way.

Cut by

Silence threshold, markers, fixed length, or a set number of pieces

Names

Timestamp, timecode or running number

Safety

Reads only — the source file is never modified

Build a name from tokens and apply it across a folder. Every proposed name is shown next to the old one before you commit, and names that would collide or contain characters a filesystem dislikes are flagged rather than silently mangled.

Tokens

{name} {project} {role} {desc} {lr} {n} {nn} {nnn}

Preview

Old and new side by side, before applying

Safety

Collisions and illegal characters flagged, never auto-fixed

Hashes the audio content rather than the file, so the same take saved twice under different names is still found. Groups what it finds and lets you move the extras to the trash — not delete them.

Compares

Audio content, not filename or timestamp

Action

Moves to trash, recoverable

Runs a folder through one conversion with one setting, writes a report of what it did to each file, and can be cancelled halfway without leaving half-written files behind.

Converts

Sample rate, bit depth, channel order, fades

Dither

Applied on the way down, off by default

Report

JSON of every file and what changed

Cancel

Stops cleanly, partial files removed

Takes a folder of stems and a master, finds the offset of each stem, sums them, and compares against the master. Writes a document you can send with the delivery — which is the difference between a client trusting your stems and a client checking them by hand.

Finds

Per-stem offset by correlation

Verifies

Sum against master, residual reported in dB

Output

A written document for the delivery folder

Generates the standard set — main, instrumental, a cappella, clean — with consistent naming and matched loudness, and packs them with their documentation into one folder ready to hand over.

Versions

Main, instrumental, a cappella, plus your own

Detection

Vocal stems recognised automatically, reviewable

Packs

Audio, stem document and metadata in one folder

Four models, each with a different opinion about where things belong. Natural keeps the front image intact and adds discreet depth. Wide pushes the sides out. Club moves energy to where a large system wants it. Fast-DA is the aggressive one.

Each model exposes what it is doing rather than hiding behind a single knob: how much bass stays in front, how much centre is extracted, how much goes to the rears and sides, and from which frequency the height layer is fed.

Models

Natural, Wide, Club, Fast-DA

Parameters

Bass keep, centre, rear, side and top amount, top high-pass

Targets

Any layout up to 9.1.6

Note

This widens an image. Lift makes objects — they are different jobs

Splits an existing stereo master into stems on your own machine — nothing is uploaded anywhere — and distributes them into the room with genre-aware presets rather than dropping them at the origin.

What comes out are real objects. You can then move them, attach triggers and export them like anything else in the scene.

Separation

Local, on your machine, nothing sent anywhere

Pipeline

11 objects out of one stereo file

Presets

13 genre-aware distributions, editable and saveable

Extras

Band split and L/R plus M/S separation per stem

A spherical microphone records the whole room at once and hands you a pile of numbered files. This turns that pile into objects sitting in the directions the capsules actually point.

Every file is measured before anything is imported and marked live, quiet, dead or clipped — a capsule silenced by a loose cable shows up here rather than three hours into the edit. You can solo any capsule while looking at where it points.

Built in

Zylia ZM-1 and ZM-1S, Eigenmike em32 and em64, OctoMic, TetraMic, Spatial Mic, A-format tetrahedral

Your own

Capsule directions from a text file or a table, saved as a reusable array

Unknown arrays

Directions estimated evenly over a sphere by the golden angle, and labelled as an estimate

Sorting

Numeric, so nineteen files land as 1–19 rather than 1, 10, 11, 12…

Placement

Rotate the whole array to match the room, or snap to the grid

Pick the interface, pick a folder, press record. Every channel is written as its own numbered file with a level meter and a clip warning while it runs. What comes out imports straight back in without renaming anything.

Channels

As many as the interface offers, checked against the array before you start

While recording

Level per channel, clip flagged and clearable

Output

One file per capsule, numbered, ready for array import

Reads filenames for role hints and analyses the signal where the name lies — a file called "audio_03" that is plainly a kick is treated as a kick. Every guess is shown with its confidence before anything is committed to the scene.

Detects

Filename patterns plus spectral and dynamic analysis

Review

Every assignment shown and editable first

Placement

Role-appropriate default positions

The single-file version of batch convert, with the BEXT metadata block written correctly on the way out — including the timecode reference, which most tools drop.

Does

Rate, depth, channel reorder, head and tail fades

Metadata

BEXT written, timecode preserved

Encodes a review version, uploads it, and gives you a link. The client opens it in a browser, listens and comments. No account for them, no large file in an email, and no master leaving your machine.

Format

Encoded review copy, never the master

Client side

A browser. No app, no sign-up

Branding

Your own artwork on the page

Sends a signal to each speaker in turn, named as you go, so a swapped pair of cables is found in thirty seconds rather than after a day of mixing into a mirrored room.

Checks

Identity, order, polarity and level per speaker

Layouts

Any rig geometry you have defined

Some platforms only accept immersive audio inside a video file. This generates a black picture at the right length and frame rate and muxes your audio under it, with an offset if the audio needs to sit late.

Picture

Black, generated at the length of the audio

Rates

23.976 to 60

Offset

Adjustable, for platforms with a fixed head

Measures the machine under load and tells you how many objects it will hold before it stutters — and writes a report you can send us when something misbehaves.

Measures

Render headroom, disk throughput, audio device latency

Output

A report file for support