Specula app icon

Workflow · Songs

The song's sections, each with its own loudness.

On macOS 27, Apple's on-device music analysis reads a song's key and tempo, its sections, bars and beats, and its four stems: vocals, drums, bass and other. Specula measures against them with its own BS.1770-4: each section's loudness under the ruler, the vocal's loudness against the bed, and the sections as chapters for a table of per-section readings.

macOS 27 Key & tempo Section loudness Bars & beats Four stems Stem-gated loudness

7-day trial · one-time purchase · music analysis on macOS 27, the rest on macOS 14+

A song in Specula after the music analysis: the sidebar's Music group with key, tempo and sections, the section band under the bars ruler shaded by loudness, and the Stems view with vocals, drums, bass and other.
A song. Key and tempo, the sections with their loudness, the bars ruler and the stems.

One number hides the parts

A track that reads −10 LUFS integrated can still run a chorus 3 LU above its verses.

Or sit its vocal too far under the bed to survive a phone speaker. Apple's analysis finds the song's structure and its stems, and Specula measures each part with the same BS.1770-4 meter as the rest of the app, so a loud chorus or a buried intro shows on the timeline, and the vocal's level against the bed is a number.

Key & tempoThe key and the BPM in the sidebar's Music group, from Apple's analysis.
SectionsEach section's integrated LUFS in a band under the ruler, shaded from darker (quieter) to lighter (louder), with the loudest, the quietest and their spread in the sidebar.
BarsA beat and bar grid: the ruler in bars, the bar and beat under the playhead, and selections that snap to a beat or a bar.
StemsVocals, drums, bass and other: where each plays and how active it is, in the Stems view.
Stem-gatedIntegrated loudness over the blocks where the vocal is present and where it is not, and the difference in LU; the same for the drums.

From file to sections

The analysis runs on its own, the timeline shows what it found, the Chapters task turns it into a table.

1 · Open the song
With the default setting, Automatically, except for speech, the analysis runs on its own after a file opens, for files up to 15 minutes that are not mostly speech; Analyze › Analyse Music runs it on any file, and Settings › Analysis also offers Only on request and Automatically, for every file. It takes about 0.6 seconds per minute of audio, plus a few seconds the first time after launch. When it finds a key, the file is a song: the sidebar's Music group opens, the ruler switches to bars for that file, the sections appear under the ruler, and the Stems view (⌃7) opens unless Settings › Analysis › Open the Stems view when a song is detected is off.
2 · Read the sections
The band under the ruler names each section with its integrated LUFS, shaded from quieter to louder, and a click on a section selects it, so ⌘Return measures it in a Selection card. The sidebar's Music group lists the loudest section, the quietest and their spread.
3 · Work in bars
The ruler's unit menu holds Time, Samples and Bars (also View › Ruler Unit). The top bar reads the bar and beat under the playhead, and the Snap chip lands selection edges on the nearest beat or bar, or stays off.
4 · The vocal against the bed
The Music group's Stem-Gated rows read the integrated loudness over the blocks where the vocal is present, over the blocks where it is not, and the difference in LU, and the same for the drums, each side needing at least 10 seconds of audio.
5 · Sections as chapters
The Chapters task (⌘4) offers a song's sections when it opens on one. From Music in its bar, or Chapters › Chapters from Music Sections, makes one chapter per section, and Analyse fills the table with each section's integrated loudness, true peak, RMS and LRA and its deviation from the song's median.
6 · Into the report
The PDF and HTML report carry a music section with the key, the tempo and the sections, and the JSON report the analysis and the stem-gated loudness.

Whose numbers

The structure, the stems, the key and the tempo are Apple's analysis and are labelled so throughout. Every loudness figure is Specula's own BS.1770-4 measurement. The analysis runs on the Mac and nothing is uploaded. It needs macOS 27; everything else in Specula runs on macOS 14 and later.

Under Podcast, Audiobook or Film and TV, a song gets a note under Deliver to, "Apple's music analysis found a key, so this looks like a song.", with a switch to Music streaming, where the delivery card also checks Apple Music's AAC encode and the mono fold-down. Mastering →

On screen

The Chapters task on a song after From Music: one chapter per section on the chapter map, and the table with each section's loudness, true peak, RMS, LRA and deviation from the song's median.
Sections as chapters. One chapter per section, each with its loudness and its deviation from the song's median.
The Stems view maximized: vocals, drums, bass and other, each lane showing where the stem plays and how active it is.
The Stems view. Vocals, drums, bass and other, where each plays and how active it is.

Part by part

On macOS 27 a song opens with its key, tempo and sections, each section carries its own loudness on the timeline, and the vocal's level against the bed is a reading in the sidebar.