MacinMeter DR Tool

English | 中文

MacinMeter measures the dynamic range (DR) of audio files. It runs entirely on your machine — nothing is uploaded, and no network connection is used.

It reads WAV, FLAC, AIFF, and Apple Lossless (.m4a / .mp4) files and reports a DR value for each channel and one for the track as a whole. There is a command-line tool and a desktop app; both give the same numbers.

Download

Get the latest build from the releases page.

Your system	File
macOS on Apple Silicon (M1 or newer), macOS 11 or later	macinmeter-gui-…-aarch64-apple-darwin.dmg
Windows, 64-bit	macinmeter-gui-…-x86_64-pc-windows-msvc-setup.exe

The macinmeter-cli-… archives contain the command-line tool on its own.

Your system will warn you the first time

This is expected. The downloads are not code-signed, so:

• macOS may refuse to open the app. Right-click it and choose Open, then confirm. You only need to do this once.
• Windows may show a blue "Windows protected your PC" screen. Click More info, then Run anyway.

Expect this to stay the case. Signing requires a certificate issued to a named individual, and that name is then embedded in every file published here. This project is maintained by one person, so signing would mean publishing their legal name alongside every download. That is a privacy cost rather than a cost in money or effort, which is why it is unlikely to change in the foreseeable future — this is a decision, not a step that has yet to be done. If you would rather not accept the trade, you can build from source instead.

You can check that a download is intact using the SHA256SUMS file on the release page.

Using the desktop app

Drag files or a folder onto the window, or use the buttons to pick them, then press Analyze. Results can be copied as Markdown or exported as JSON, PNG, or SVG. The interface is available in English and Chinese.

Using the command line

mdrmeter analyze "01 - Song.flac"
mdrmeter batch "My Album/" --recursive

batch always lists files in the order you gave them, whatever order they finish in, and one unreadable file does not stop the rest. It reports each track on its own and does not compute an album DR.

Real output from a test file included in this repository:

MacinMeter
Source: tests/fixtures/edge_cases.wav
PCM: 44100 Hz, 2 channels, 308700 frames
Duration: 0:07

CH 1: DR2 (2.4300 dB), overall RMS -2.43 dBFS, selected DR peak 1.00000000
CH 2: DR2 (2.4300 dB), overall RMS -2.43 dBFS, selected DR peak 1.00000000

Track aggregate: DR2 (2.4300 dB; 2 contributing channels)
Report levels: peak 0.00 dBFS, RMS -2.43 dBFS

Elapsed: 0.002 s (2929.6x realtime)

That file is a synthetic test fixture, not an example of a real music release.

Useful options:

Option	Effect
--format json	machine-readable output instead of text
--output PATH	write the report to a file instead of the screen
--recursive	with batch, also search inside subfolders
--timing	also show how long decoding and analysis each took

Shell completions are available for bash, zsh, fish, PowerShell, and elvish:

mdrmeter completions zsh > "${fpath[1]}/_mdrmeter"

Reading the result

• DR2 is the track's DR value. A larger number means a larger gap between the peaks and the loud parts' average level. A high DR value does not by itself mean a recording sounds good, but a very low one often indicates heavy compression and a more likely compromised master. Genre and artistic intent still matter.
• Each CH line is one channel's own DR, its overall RMS, and the peak the DR calculation selected.
• Report levels are whole-track measurements. The peak reported there is a different quantity from the selected DR peak above.
• dBFS treats amplitude 1.0 as 0 dB. Floating-point audio can legitimately contain samples above that, so 0 dBFS is not always a clipping point.
• Silent channels stay visible and count as DR0. Channels with too little data to measure are excluded and said to be excluded.
• Elapsed and the realtime multiple describe this run on this machine, not the analysis itself, so they appear only in the text output. JSON stays identical between runs of the same file.

Some reports add warnings — for example when a file is shorter than one analysis window, or when a multichannel file's speaker layout is unknown and the track value may therefore include an LFE channel. These never change the numbers.

Exit codes

Code	Meaning
0	everything analyzed successfully
1	failure, no input, every batch item failed, or the report could not be written
2	invalid command-line arguments
3	batch finished with both successes and failures
130	cancelled

Which files work

Format	Supported
WAV	8/16/24/32-bit integer, 32/64-bit floating point
FLAC	native FLAC files
AIFF	8/16/24/32-bit integer
Apple Lossless in .m4a / .mp4	16/24-bit, 1–8 channels

Folder scans look for .wav, .wave, .flac, .aif, .aiff, .m4a, and .mp4. A supported file with some other extension still works if you pass its path directly, because files are identified by content rather than by name.

Not every file with these extensions will work. AAC, MP3, Ogg, Opus, and DSD are not supported at all, and some less common variants of the formats above are not either. MacinMeter says so plainly instead of guessing: it never converts, resamples, or falls back to another decoder. The format guide lists the exact rules and every excluded variant.

How accurate is it

The DR algorithm was reconstructed from one specific program, foo_dr_meter 1.0.8 x64, with the original author's permission. On a fixed set of recorded test inputs, MacinMeter's results match that program's exactly — 39 of 39 tracks and 62 of 62 channels, with no differences.

That is a precise claim about a specific test set, not a promise about every file in the world. What was compared, and what was not, is listed in the accuracy record.

More

• Format guide — exactly which files work and why
• Technical notes — architecture, Rust API, accuracy evidence, performance measurements
• Release and packaging status
• Legal notes and third-party notices

MacinMeter is released under the MIT License.

The reference target is Janne Hyvärinen's foo_dr_meter 1.0.8 x64 component. Reverse-engineering it was done with the author's permission; only a minimal public authorization summary is kept in this repository.