A free update for every Specula 1.x license. 2.0 organizes the window around seven tasks and Deliver to. Where 1.4's Mode, Dialogue mode, Chapter mode and their settings went is listed in the user guide's Coming from 1.4.
- NewOne window, seven tasks. The window is organized around seven tasks: Inspect, Edit, Compare, Chapters, Speech, Spectral and Record, chosen in the top bar or with ⌘1 to ⌘7. Each task brings its own bar of tools, its own views and its own part of the sidebar, and its shortcut, or Esc with nothing selected, returns to Inspect. The Speech task replaces Dialogue mode, the Chapters task replaces Chapter mode, and Deliver to replaces the Mode.
- NewInclude harmonics in Spectral. A switch in the Spectral edit panel and the Spectral menu adds the selection's tone at 2×, 3× and up, to 16× (8× to start), to what Apply, Solo and Hear Result edit: a squeak, a whistle, feedback or hum with its overtones. Each time column is read on its own, so the harmonics of a gliding pitch follow it. Only where they ring, on to start, takes a harmonic only where its peak stands 12 dB above the audio around it, and only as wide as it rings; off, each harmonic takes the selection's band times its number. The canvas outlines the harmonics with the selection, and the panel names the ones found.
- NewThe top bar. One bar holds the sidebar toggle, Open, the transport (Stop, Play, Loop, Hold play start, and Record in the Record task), the playhead time with the bar and beat, the playback speed and pitch, IN, OUT and LEN with Clear and Analyse, the time zoom and snap, the tasks, the task's views with Maximize, the output device and Export. In a narrower window its groups wrap onto a second row, the transport staying on the first. The seek bar runs under it.
- NewViews for each task. Inspect opens on the waveform, loudness and spectrogram, Edit and Speech on the waveform and loudness, Chapters on the chapter map over the chapter table, Compare on the metrics matrix over the waveform, Spectral on its canvas and Record on its meters and the last take. The view buttons in the top bar and ⌃1 to ⌃7 show and hide views, and each task remembers its own; Maximize, or ⌃⌥1 to ⌃⌥7, lets one view fill the canvas until it is restored. A drag on the line between two views shares out their space, and a double-click resets it. Zoom and scroll stay when the task changes.
- NewEdit. The Edit task's bar holds Cut, Trim, Fade In, Fade Out and Gain, a Loudness menu (Normalize, Limit True Peak, Level Dialogue), a More menu (Silence, Invert Phase, Remove DC Offset, Swap Left and Right, Split to Mono Files), the Plugins drawer, Save and Save As. A selection shows its length over the waveform with the common edits beside it. The sidebar's History lists every edit by name, up to 16 steps back; a click on a step goes back or forward to it in one move, and the step that matches the file on disk says On disk.
- NewCompare. The Compare task shows every slot's readings in a matrix over the waveform, each with its difference from A, a Residual row and a Listen button per slot (1 to 6). Its bar holds the slot chips, dragged to reorder, with a menu to make a slot the reference, move it or remove it, then Add, Align, Match, Ghost, Residual and Export Residual. The sidebar's slot card holds the slot's offset with nudges and Align, its level match, its trim and its polarity. Moving a file into slot A rebases every alignment onto it.
- NewChapters. The Chapters task's bar holds Detect, From Music, Add, Remove, Analyse, the detection settings, Recall, Import, Export, Export CSV and Save. The chapter table lists every chapter's readings, with ACX's marks under Audiobook and Podcast, and the sidebar holds an ACX summary and Level chapters with each chapter's gain before Apply. Opened on a file without chapters, Chapters detects and measures them, or offers the chapters saved in an earlier session, or a song's sections.
- NewSpeech. The Speech task's bar holds Add, In, Out, Split, Delete, Reset to Detected, the detection settings (moved from Settings), Export and Import of the regions, and Save. The sidebar's Level dialogue panel shows the dialogue loudness and noise floor now, a prediction of both after leveling, and the floor against ACX. Undo and Redo in Speech step through its region edits and Level Dialogue's Apply, newest first. A detection setting never replaces corrected or imported regions, and Detect Again asks first.
- NewSpectral. The tools sit in a column on the canvas's left edge (Rectangle R, Lasso L, Brush B, Wand W), the operation is a panel in the sidebar (Attenuate, Erase or Repair with its amount, the feather, Include harmonics, the channels it edits, Solo, Hear Result and Apply), and the task bar holds the tool's options, Clear, the analysis size, the spectrogram's Floor and Ceiling knobs, the channels shown and Save. Hear Result plays the chosen operation as Apply would write it.
- NewRecord. The Record task's canvas shows each input channel's level and peak, the momentary loudness and the true peak, and while armed, a room check: the room's floor, measured as the file's noise floor will be, against ACX's limit. The sidebar holds the input (device, layout and source channels), the room check, and the takes of the last seven days with To Compare, Save As and Move to Trash. The bar chooses where a take goes: a new file, Append, Insert at the playhead or Punch in.
- NewThe empty window. With no file open, the window offers Open, Generate, Record and Check a Folder, and lists the ten most recent files with their length, format and the delivery category they were last judged against. When no delivery category was ever chosen, it asks once where audio usually goes, which sets Deliver to.
- NewA drag selects. In every lane a click places the playhead and a drag selects, and a ⌃-drag moves the playhead. Settings › General › Pointer offers 1.4's way instead: a drag moves the playhead and a ⌘-drag selects. A pinch zooms time, a sideways swipe pans, a vertical scroll zooms and Shift pans, and every lane zooms to the same depth.
- NewReports say what they were judged against. The PDF and HTML report name the delivery category (Music streaming, Podcast, Audiobook, Film and TV or Broadcast) and list its targets with the primary one first: "Judged against: Spotify (primary), Apple Music, YouTube, Tidal". The JSON report carries the same in a
delivery block: the category, the primary target and the checked targets.
- NewMusic analysis skips speech. With the default setting, Apple's music analysis runs on its own only for files that are not mostly speech; Analyze › Analyse Music still runs it on any file. The settings name what each option does: Only on request, Automatically, except for speech, and Automatically, for every file.
- NewSong sections in one place, shaded by loudness. The section band under the waveform's ruler shades each section from darker (quieter) to lighter (louder), with no red outlier marks, and while the band shows the sections, the Music module leaves out its own structure row. The Music module's pace row reads in events per minute.
- NewA song opens its Music group and bars. When the music analysis finds a song, the sidebar's Music group opens and the ruler switches to bars for that file; the next file returns to the ruler's previous unit, and a unit chosen while the song is open stays. The Stems view (⌃7) opens too, unless Settings › Analysis › Open the Stems view when a song is detected is off.
- NewThe loudness graph opens on the file and its target. The loudness section opens on a range around the file's loudness and the primary delivery target, and draws the target as a dashed line with its name and level, such as Spotify -14 LUFS. It used to open on -60 to 0 LUFS, which squeezed a master's curves into the top of the graph. An Option-scroll zoom stays until you double-click the graph, open another file or choose another target. ACX targets get no line.
- NewReadings at rest, and live ones on request. The sidebar shows the whole file's readings and target verdicts as soon as the file has loaded, and they hold still while the file plays. The Whole file and Live switch on Loudness details shows what is playing instead: momentary and short-term loudness, and the integrated loudness, loudness range and true peak of what has played since playback was stopped, the file was edited or another slot was chosen, with short bars. The switch starts on Whole file, and the target verdicts always judge the whole file. In 1.4 the sidebar started on Live and read -∞ until playback started.
- NewReadings follow every edit. After an edit, Undo or Redo, the file is measured again in the background, and the sidebar and the target verdicts show the edited file. They used to stay empty until ⌘Return.
- NewReadings follow corrected speech regions. Correcting a speech region in the Speech task, reopening a file with saved corrections or changing the speech detection settings measures the file again, so the dialogue loudness, Speech % and noise floor, the target verdicts and fixes and the reports follow the corrected regions.
- NewThe spectrogram shows the whole file. It is computed when the spectrogram section is shown, with its progress in the lane, and stays as it is while the file plays; ⌘Return no longer replaces it with the selection's. Settings' Auto-compute spectrogram on load is gone, and no Live setting changes the spectrogram. On a multichannel file the lane names the channel it shows.
- NewThe spectrum and phase panels show the whole file at rest. When a file opens, the spectrum shows the whole file's average spectrum, labelled Whole file. Placing the playhead shows the spectrum there, labelled with its time, and after ⌘Return the panel shows the analysed selection's while that range is still selected. The label is a menu for switching between them. While playing the spectrum is live, as before. The phase panel plots where the whole file's left and right samples fall, with their correlation. A section with nothing to show says why, such as "Mono file: no stereo width".
- NewThe sidebar shows what fits the file. The Speech group (dialogue loudness, Speech % and noise floor) appears when the file's Speech % is 10 % or more, or when the primary target judges speech (dialogue loudness, or ACX's RMS and noise floor). It no longer depends on the Mode: a narration checked against Spotify shows its dialogue loudness and Speech %, and an instrumental checked against EBU R128 does not. The Noise Floor row appears for Podcast and Audiobook delivery and for any target that checks the noise floor. The Music group (Key, Tempo, Sections and Stem-Gated) appears once the music analysis has found a key. The loudness section's Dialogue overlay and SG-I line appear with the Speech group, and the PDF and HTML reports' Speech-Gated and Noise Floor cards follow the same rule.
- NewLevel Dialogue only on files with dialogue. Level Dialogue is unavailable until speech detection has finished, on a file whose Speech % is under 10 %, and, unless the primary target judges speech, on a song (the music analysis found a key) or on a file the music analysis has not checked. Its tooltip gives the reason.
- NewTarget fixes in words. The three icons on each loudness target row are replaced by words. The primary target's fixes are buttons that say what they do, such as Limit peaks to -1.0 dBTP, Normalize to -14 LUFS and Hear at Spotify's level, and every other target has one Fix menu with its own. A fix appears only where it changes something: Limit when the whole file's true peak is over the target's ceiling, Normalize and Hear when the target would change the file's level. A file within a target's tolerance, or a quiet master that Apple Music, YouTube, Tidal or Apple Podcasts play as recorded, gets no level fix. With more than one file in Compare, only Hear is offered.
- NewDeliver to. Over the delivery card the sidebar names what the file is judged against, such as Deliver to Spotify · Music streaming. Its picker lists the five categories (Music streaming, Podcast, Audiobook, Film and TV and Broadcast) and each category's targets, with a checkbox for each and a choice of the target the verdict leads with; custom targets appear in their category, and Edit targets… opens Settings. Analyze › Deliver To… (⇧⌘D) opens the picker from the keyboard. It replaces 1.4's Mode section. Audiobook is new: Podcast with ACX leading, so a Podcast list that led with ACX in 1.4 opens as Audiobook.
- NewLens suggestions. When the file fits another category better, a note under Deliver to says so and offers the switch, such as This file is mostly speech. Switch to Podcast. It covers speech under Music streaming, a song under Podcast, Audiobook or Film and TV, a surround mix under Music streaming and a long chaptered narration. The lens changes only when the switch is pressed, and the note's close button hides it for that file until Specula quits; the Deliver to picker still shows it. It replaces the Song detected button.
- NewThe Normalize panel shows Now, Target and Gain. Now is the sample peak in Peak mode and the integrated loudness in LUFS mode, of the selection or of the whole file, whichever Normalize applies to. A selection that ⌘Return has not analysed shows no Now and asks for ⌘Return.
- NewThe sidebar starts with the file and the task. The file's name heads the sidebar, with its format, sample rate, encoding, channels and length under it; its size and location open from there. In every task but Inspect the task's own tools come next, under the task's name in its colour: the edit history in Edit, the slot in front in Compare, ACX and Level chapters in Chapters, Level dialogue in Speech, the spectral edit in Spectral, and the input, room check and takes in Record. Deliver to and the delivery card follow, then Loudness details, Speech, Music and Channels.
- NewA delivery card in the sidebar. Under Deliver to, a card names the primary target and gives its verdict in a sentence, such as "Spotify turns this down by 5.0 dB." or "Apple Podcasts plays this as recorded, 8.7 dB under its -16 LKFS reference.", with the file's numbers in the target's own terms (integrated loudness, dialogue loudness, or RMS and noise floor for ACX) and the target's fixes. The other checked targets follow, each with its short verdict and a Fix menu where a fix applies. The card always judges the whole file.
- NewA selection's readings in their own card. After ⌘Return, the analysed selection's readings show in a Selection card below the delivery card, with the time range, a button to clear them and the per-channel readings on request. The target verdicts keep judging the whole file.
- NewOne set of colours. A reading is coloured by what the primary target does with it: green within the target, within 1.5 dB of a platform's reference or where the platform plays the file as recorded, yellow up to 4 dB away, orange beyond that or for true peak over a streaming platform's ceiling, red where a delivery spec fails. Readings no target judges, such as sample peak, loudness range and max momentary, are not coloured. The sidebar, the per-channel statistics window and the PDF and HTML report use the same colours, the report from its primary target.
- NewThe sidebar's groups remember their state. Loudness details, Speech, Music and Channels open and close from their headers, and the file's size and location from its name, and stay that way across launches; the sidebar keeps its width.
- NewPRE and POST are in the Output panel too. The switch between the file's channels and the routed output for the sidebar's channel meters also sits next to the routing matrix, and the Channels group names what its meters show.
- NewOne export sheet. File › Export… (⌘E) writes the PDF, HTML and JSON reports, the chapters CSV, the edited audio and a compare residual into one folder in one step, named after the file. ⇧⌘E opens it with the JSON report ticked and ⌥⌘E with the PDF report. What the reports include (loudness curves, stereo width, per-channel statistics, violations) is set in the sheet and remembered for each delivery category, starting from the Settings › Report choices; that Settings tab is gone. The sheet shows the delivery category and the targets the reports state. An existing file is replaced only after asking, with Keep Both as the default. Right after an edit, Export waits until the file has been measured, so the reports carry its readings. The report preview opens from Window › Report Preview. HTML reports are new in the app.
- NewCheck a Folder. File › Check a Folder…, or the Check a folder tile with no file open, checks every audio file in a folder against Deliver to, with Change… beside Judged against to pick the delivery category and targets. Each file gets a row with its verdict, integrated loudness, true peak, loudness range and Speech %, measured with the app's speech settings and any saved speech regions made for it. JSON, HTML or PDF reports are written for each file as it finishes, and a CSV of the rows at the end, by default into a new folder beside the checked one, never into it and never over an existing file. Subfolders are included on request. After a check, a new target judges the rows again from their readings; the reports and the CSV keep the target they were written with.
- NewThe report's chapter table checks ACX. The PDF and HTML report group the chapter columns as ACX (RMS, true peak, noise floor), Consistency (integrated loudness against the book's median) and Details (LRA, max short-term, dialogue loudness, speech). Under Audiobook and Podcast each ACX cell is green when it passes and red when it fails, with ACX's limit as a tooltip in the HTML report; under the other lenses the cells show their readings without a verdict, and a value just past a limit prints with the decimals that show it, such as -17.96 rather than -18.0. The sample peak and max momentary columns moved to the chapters CSV only.
- New
specula report takes the speech detection settings. --vad-threshold, --vad-min-speech and --vad-min-silence set the threshold, the shortest speech region and the merge gap, within the ranges of the app's settings, so a report can measure speech with the same settings as the app.
- NewReports say what speech was measured with. The PDF and HTML report state whether speech was detected, with the detection settings when they are not the defaults, or came from corrected speech regions, and the JSON report carries the same in
fullFile.speechMap. specula analyze and specula compare add speech.map (detected, corrected or none), and the report CSV a Speech map row.
- NewSettings in five tabs. General (appearance, text size, Icons only for the top bar's task and view buttons, the selection gesture, the loudness readout, updates, the license and a button that installs the command-line tool), Analysis (the FFT window, the spectrogram and music analysis), Delivery targets (each category's targets and the custom targets), Editing and plugins, and About. Speech detection is set in the Speech task, chapter detection in the Chapters task and the report contents in the export sheet. Each tab scrolls. The Delivery targets lists keep the Deliver to picker's rules: under Audiobook ACX stays checked, and under Podcast unchecking the target the verdict leads with passes the lead to another target, never to ACX.
- NewVoiceOver names every control. Buttons, fields, switches, pickers and sliders read a name that says what they do, and every choice says which option is chosen: in the Signal Generator, the Output panel, the raw import sheet, the custom target editor, the editor of a plugin without its own view, the plugin chips, the zoom and close buttons, the gesture legends and the license key field. Each routing matrix cell is a button named for its route, such as L to R Out, with its state, and its actions cycle the phase and set the gain. The tiles shown with no file open and the chapter map's chapters are buttons, and a chapter's actions include Rename and Jump playhead to start.
- NewThe keyboard reaches every control outside the main window. In the Signal Generator, the Output panel, the report preview, the statistics window and every popover, Tab moves to each control and Space presses it. In the Output panel a gain takes Return or Space to type a value and the arrow keys to step it by 1 dB. With no file open, Tab reaches the main window's tiles and Space presses them.
- NewEvery sidebar and drawer action is in the menus. In the main window, where Space plays, the menus reach what the sidebar's panels do: Chapters › Level Chapters (⇧⌘L) and Speech › Level Dialogue apply the panel's settings, and Spectral › Attenuate applies the panel's gain, which the item names. The Edit menu adds a plugin, restores the last chain, renders it and cancels a render, and acts on the selected plugin: Select Next Plugin, Bypass, Move Earlier or Later, Show Controls and Remove. The Record menu acts on the selected take: Select Next Take, Add to Compare, Save As and Move to Trash. A click on a plugin's or a take's name selects it.
- NewInspect is ⌘1. Edit, Compare, Chapters, Speech, Spectral and Record follow on ⌘2 to ⌘7, in the order of the task buttons. In 1.4, ⌘1 to ⌘6 were Edit to Record.
- NewSpectral selections are outlined. A selection on the spectral canvas has a white outline and keeps its own colours, the rest of the canvas is dimmed while it exists, and the feather's roll-off shows as a soft edge around it. Its time span and frequency range print beside it. The selection was a faint green wash with no edge.
- NewContrast in Spectral. Spectral › Contrast… and the Contrast button in the task bar set the level the canvas draws black and the level it draws brightest, so a faint blemish can be brought into view. The spectrogram section uses the same range.
- NewHDR accents. On macOS 26 and later, on a display that shows HDR, a few things light brighter than white: the playback point on the seek bar while playing, the record light while a take rolls, a clip or an over on the meters, the brightest spots of the live phase scope and the outline of a Spectral selection. They light only while Specula is the app in front, and on a display such as Studio Display only below full brightness. Otherwise they keep their usual colours.
- NewSpectral's rectangle and brush. The rectangle's outline shows the cells it will take while it is dragged, which at the bottom of the frequency axis can be taller than the drag. The brush's radius starts at 12 points, where it was 40.
- NewA chapter click selects the chapter. Clicking a chapter in the chapter map makes its time range the selection, in every task that shows the map, as a row of the Chapters table does. Space then plays that chapter.
- NewFit for the amplitude zoom. The amplitude zoom under the waveform has the time zoom's buttons: zoom out, zoom in and fit, which sizes the loudest peak of the unmuted channels to the lane. Double-clicking its readout returns it to full scale, and Option-scroll zooms it smoothly.
- NewThe ruler names its unit. The label at the end of the ruler shows the unit the ruler is in (Time, Samples or Bars) and opens a menu of the three, the current one checked; in 1.4 it named the unit a click would switch to. Bars stay in the list, greyed out, until the music analysis has found a beat grid. View › Ruler Unit holds the same three.
- NewThe codec check. Analyze › Check the Apple Music Encode, and a button on the delivery card under Music streaming, encode the file as AAC, the same encode Apple's afconvert makes at 256 kbps, constrained VBR and quality 127 (128 kbps for mono), decode it and measure the encode's true peak and its samples over full scale. Files above 48 kHz are resampled to 44.1 or 48 kHz first. The card states the result, such as "After Apple Music's AAC encode the true peak is +0.4 dBTP, with 12 samples over full scale." When the encode is over the primary target's true-peak ceiling, Limit peaks opens Limit True Peak at the target less what the encode added. Hear the encode in Compare, or Analyze › Compare with the Apple Music Encode, adds the encode as a Compare slot beside the master, where its residual (D) is what the encoder removed. After an edit the result reads as checked before the last edit until the file is checked again. The check takes mono and stereo files at 44.1 kHz and above.
- NewThe codec check in the reports. The export sheet's Codec check switch adds the check to the PDF, HTML and JSON reports: the master beside the encode (integrated loudness, true peak, sample peak and samples over full scale), what the encode changed, the ceiling it is judged against and the limit that brings it under. The switch is remembered for each delivery category and starts on for Music streaming. When the file has not been checked as it is, Export runs the check first.
- NewThe mono check. Under Music streaming, the delivery card says whether a stereo file folds to mono. It reads left against right as the Stereo width view does, over the whole file and over every second of it, and a file passes when its correlation stays above -0.2, such as "Mono-compatible. Correlation +0.62 over the file." Where some seconds fall below, the card says how long and from where, and Select the longest stretch selects that stretch to play it. A whole file below -0.2 reads out of phase, as a file with one channel's polarity flipped does.
- NewThe loudness thresholds name their units. The M, ST and TP fields under the loudness graph show LUFS and dBTP beside them, and their tooltips say what each one marks and the primary target's own limit where it has one, such as Spotify's -1 dBTP ceiling. The fields start empty, which marks nothing.
- NewBrackets for In and Out, Generate on the right. Speech's In and Out buttons show [ and ] instead of play-style triangles, and the Signal Generator's Generate button sits at the bottom right of its window.
- NewProgress and Cancel while a file opens. Opening a file shows its name and how far the decode has got, with a Cancel button; Esc cancels too. The file already open stays as it was until the new one has decoded, so a cancelled open leaves it untouched, with its edits, its speech regions and its plugin chain. Nothing plays while a file opens.
- NewKnobs for the spectrogram's range. A strip under the spectrogram section holds two knobs, Floor (the level drawn black) and Ceiling (the level drawn brightest), with the section's gesture legend beside them. Drag one up or down to turn it, with Option for fine steps, and double-click it for its default. In Spectral the same knobs sit on the task bar in place of the Contrast button; Spectral › Contrast… still opens the sliders.
- NewACX in its own words, its limits as it states them. ACX asks for a volume between -23 and -18 dB RMS, peaks less than -3 dB and a noise floor less than -60 dB RMS, and Specula now judges and words them so: a true peak of exactly -3 dBTP or a noise floor of exactly -60 dB RMS fails, where 1.4 passed both. Limit peaks for ACX limits to -3.1 dBTP, and Level Chapters starts there. Every other target's ceiling is still a maximum the limit itself meets.
- FixOne playback at a time in Spectral. Pressing Space while Solo or Hear Result played started the file on top of it. Space now stops the audition and starts nothing, Play and Stop end it too, and starting Solo or Hear Result pauses the file. While an audition plays, the Play button shows pause. Solo is ⇧Space and Hear Result ⌥Space.
- FixTarget fixes work from the whole file. Normalize to a dialog-gated target (Apple Podcasts, Spotify podcasts, the film and TV targets, ATSC A/85), Level Dialogue's lift to a target and Match playback took their gain from the sidebar's reading. With the factory Live setting that was the dialogue loudness of whatever had played so far, and after ⌘Return on part of the file it was that part's, so a normalized file could land away from the target. They now work from the whole file's reading, and the Normalize panel's Now reads "whole file". After a switch to another compare slot, Match playback works from that slot's reading.
- FixCompare slots fill their lanes when their measurements land. A slot brought to the front before its measurements had finished kept empty loudness and stereo lanes until the next switch.
- FixAn early edit keeps its own readings. Editing a file before the analysis that runs on opening had finished put the unedited file's readings back in the sidebar when that analysis ended, and the target fixes then worked from them.
- FixSave As never replaces the source under another spelling. On a case-insensitive disk, the macOS default, a Save As typed with other capitals (Song.WAV for Song.wav) could replace the source file once the save panel's replace question was confirmed. Save As, Export Diff, Split to Mono, rendering the routed output, the Signal Generator and
specula edit now refuse any name that opens the source file, a hard link included.
- FixChapter names are remembered for files with a silent chapter. When a chapter was pure digital silence, Specula could not save the file's chapters for the next session, so renamed chapters came back with their old names, and exporting the chapters as JSON failed. Both work now. Chapter JSON writes a silent chapter's readings as "-inf", as the JSON report does.
- FixReports in Compare describe the file in front. With a slot other than A in front, a report took that file's name, loudness and sample peak but slot A's true peak, loudness range and max momentary and short-term, so its verdicts could judge a mix of the two files. A report now describes the slot in front, all of it, as the sidebar does.
- FixSave As asks before it clips. An edit can leave samples above full scale, which a 16-bit or 24-bit file cannot hold, and Save As clipped them without a word while Specula kept playing and measuring the unclipped audio. When the saved file would clip, Save As now asks first and offers 32-bit float, which keeps every sample.
- FixThe window stops before it cuts off its controls. The window could be dragged narrower than its controls, down to the sidebar alone. It now stops where every bar has wrapped as far as it goes, and the controls under the loudness and spectrum sections wrap to a second row where they do not fit.
- FixThe Size row updates on save. After Save As, the Size row read 0.0 MB until the file was opened again, and a report exported meanwhile gave the file's size as 0. After Save over the file, it kept the size the file had when it was opened. Both saves now read the size of the file they wrote.
- FixReports judge the whole file. In the app, a report's target verdicts took RMS and the noise floor from the analysed selection only, so a report exported straight after opening a file read ACX as unavailable, and after ⌘Return on part of the file ACX's verdict mixed that part's RMS and noise floor with the whole file's true peak.
specula report, the Get Report shortcut and the Finder's Get Report service did not check ACX's noise floor at all. Every report now judges the whole file, noise floor included, and the JSON report's full-file block carries the noise floor. With a selection analysed, the PDF and HTML report say their readings are the selection's and that the targets judge the whole file.
- FixReports give the same Speech % as the sidebar. Reports worked out Speech % over the whole file, silent stretches included, so a narration with silent pauses read lower in its report than in the sidebar, in
specula analyze and in the Get Measurements shortcut. Reports from the app, specula report, the Get Report shortcut and the Finder's Get Report service now give the same number.
- FixLeaving Edit asks about unrendered plugins on every path. Switching from Edit straight to another mode set an unrendered plugin chain aside without asking. It now asks whether to render the chain into the audio or discard it, as the Done button always did.
- FixThe spectrogram never shows the previous file. After opening another file, the spectrogram could keep showing the previous file's picture until the new one was computed, which took minutes for a long narration. It now clears at once and says "Computing spectrogram…" until the new file's columns are in, including for two versions of a song with exactly the same length.
- FixA short file's spectrogram is smooth. The spectrogram drew each analysis window as a flat block, so a few seconds of dialogue read as a grid of rectangles beside the same audio's smooth picture in Spectral. It now blends between neighbouring windows, as Spectral does.
- FixThe spectrogram lines up with the playhead. Each spectrogram column now sits at the centre of the audio it analyses, so a drum hit's stripe lines up with the waveform and the playhead in every section. It used to draw half an analysis window off, about 85 ms early at the default settings, or late while playing.
- FixShortcuts work after closing a popover. Closing an Edit popover, such as Normalize or Level Dialogue, could leave the keyboard in the selection's IN field, so Space, L and the number keys typed into it and Return set a stray selection. The IN, OUT and LEN fields, the Compare task's offset and gain fields and the loudness section's M, ST and TP fields now take the keyboard only when clicked or reached with Tab. Clicking into one of them and out again leaves its value as it was, and Esc in IN, OUT or LEN puts the value back and hands the keys back.
- FixEach compare slot keeps its own spectrogram. With the spectrogram computed on load, switching slots while it was still computing drew it in the other slot's lane, and it stayed there. It now waits in its own slot.
- FixThe Analyse buttons work after opening a file mid-analysis. Opening another file while an analysis ran left the Analyse buttons disabled and the spectrogram saying "Computing spectrogram…" until another analysis, started with ⌘Return, had finished.
- FixEsc does one thing per press. One press used to clear the waveform selection and the FFT band together, and in Spectral what it cleared depended on timing. Each press now clears one thing, in a fixed order: a drag in progress, then the waveform selection, then the FFT band, then the selected chapter or speech region; in Spectral, the selection and its span together. With nothing left to clear, Esc goes back to Inspect. It never ends a recording in progress. In the Signal Generator, Settings and report windows, Esc no longer clears the main window's selection.
- FixHeadroom is measured from the true peak. The Headroom reading used the sample peak, so a master with inter-sample overs showed room to spare. It now reads from the true peak, in the sidebar, in compare slots and in reports, and goes negative when the true peak is above 0 dBTP.
- FixOpening several files fills the compare slots in file-name order. Opening Song v1 to v4 together could fill the slots as v1, v3, v2, v4, depending on which file finished loading first. The slots now follow the files' names in the Finder's order ("Take 2" before "Take 10"), and the first becomes slot A. Add Files to Compare follows the same order.
- FixA residual shows its own readings. While you listen to the difference between A and another slot, the sidebar shows the residual's own peak and RMS; it used to show the other slot's loudness under the residual's name. A residual gets no loudness or target verdict.
- FixExport Residual is available only when there is a residual to export. In 1.4 the Export Diff button looked enabled when it was not, and the Export Diff Audio menu command exported another slot's residual when the active slot had none. The Compare task bar's Export Residual and File › Export Residual Audio… export the active slot's residual, and disabled buttons look disabled.
- FixA new slot A keeps the other slots aligned. Moving a file into slot A, or removing slot A, left every other slot's offset measured against the previous A, so their residuals no longer nulled and the Waveform A Overlay was drawn out of place. The offsets now follow the new A, the residuals are computed again against it, and a note names the new reference. A file moved into slot A while its residual played kept playing that residual; it now plays the file.
- FixAuto-align updates the residual. After Auto-align, the slot's residual stayed computed at the previous offset until the offset or gain was changed by hand. It now follows the new offset, and an alignment still running when the slots are reordered lands on the slot it was measured for. A change to a slot's offset, gain, polarity or level match made while its residual is still being computed, such as Auto-align followed at once by listening to the residual, reaches the residual too: it is computed again with the slot's settings.
- FixA punch-in at the playhead keeps the file's length and the speech regions around it. Each punch at the playhead shortened the file by both splice crossfades, 20 ms at the default setting, moved everything after it earlier, and dropped the chapters and any speech regions corrected by hand. The crossfades now lie inside the punched stretch, so the file keeps its length and everything after the punch stays where it was. Speech is detected again inside the punched stretch, also after Undo and Redo across a punch, and corrected regions elsewhere stay as they are. A take appended to the end, or a punch that runs past the end, keeps the chapters and the corrected speech regions too, and speech is detected in the new stretch.
- FixSave Take As… saves the take. In 1.4, Record mode's Save Take As… saved the loaded audio, which after an Append was the whole file with the take in it. It now saves the last take, as 32-bit float in WAV, AIFF or CAF.
- FixThe room check reads the room as the file will be measured. While the input is armed, the room's floor over the last five seconds is read the way Specula measures a file's noise floor (the mono mix, the 90th percentile of its 100 ms blocks) and judged against ACX's noise floor limit. It used to show the quietest moment of those five seconds, so a room that hummed or breathed read several dB better than the recording would measure. An input that delivers digital silence, such as a microphone without permission, reads No input signal instead of passing.
- FixSpace and the slot keys act only in the main window. With the Signal Generator or Settings in front, Space could start playback in the main window, and in Compare 1 to 6 could switch its slot. They now act only in the main window, and the Play / Pause and Switch to Slot menu items name their keys in their titles.
- FixDropping several files on the empty window opens them all. A drop on the drop zone shown with no file opened the first file and left the rest; they now all open, as a drop elsewhere on the window did, the first as the file and the others in Compare.
- FixCorrected speech regions count on the command line and in Shortcuts.
specula analyze, compare and report, the Get Measurements, Get Report and Compare Files shortcuts and the Finder's Analyze, Compare and Get Report services measured a file with freshly detected speech and ignored its .dlg.json sidecar, so a file with corrected speech regions read a different dialogue loudness, Speech % and noise floor there than in the app. They now use the sidecar's regions when it was made for the file's length, as the app does, and --no-sidecar measures with the detected speech.
- Fix
specula compare --out-diff is consistent for identical files and always writes WAV. For two identical files the residual's RMS read -200 dB while its peak was left out; now both are left out. A residual saved under a name like Residual.WAV came out as a CAF file; it is a WAV file now.
- FixJSON reports export for files with digital silence. A file with a stretch of digital silence, such as a pause of pure zeros or silence at the start, could not be exported as a JSON report from the app or from Shortcuts: the export stopped with an error. Silent stretches in the loudness curves now read "-inf", as the command-line tool already wrote them.
- FixThe report and the sidebar say "digital silence" for a silent noise floor. When the pauses were digital silence, the PDF and HTML report and the sidebar's Noise Floor row showed -200.0 dB as the noise floor, which read like a measurement. They now say digital silence. The JSON report keeps the number.
- FixThe PDF report's chapter table fits the page. The chapter table was wider than the page, so its last columns were cut off in the PDF. It now fits, long chapter names wrap, and the time range reads "12:30 to 15:00".
- FixThe report's loudness curves break at pauses. The momentary, short-term and integrated curves drew a straight line across every stretch of digital silence. They now stop at the pause and start again after it.
- FixRegion edits made just before leaving the Speech task are saved. In 1.4 an edit made less than half a second before leaving Dialogue mode never reached the
.dlg.json sidecar, so it was gone the next time the file opened. Leaving the Speech task now writes it.
- FixSilence reads as no reading. A silent channel's sample peak, crest factor and DR read "-", and an RMS at the -120 dBFS floor reads "below -120", in the sidebar's Channels group, the per-channel statistics window and the PDF and HTML report, where they read -240.00 dBFS, -120.00 dB, DR1 and -120.00 dBFS, the lower limits of the measurement rather than readings of the audio. A silent file's true peak and DR read "-" on the report's cards as in the sidebar, and under ACX its RMS reads "below -120 dB RMS", in the verdict too. A silent selection's sample peak reads as the whole file's does. The JSON report keeps the numbers, except a silent analysed selection's
selection.samplePeakDB, which is "-inf" as the whole file's is, where it was -240.
- FixNumbers print the same in every region. Values print with a point for decimals and no thousands separator whatever the system's region, as the reports and the command-line tool print them: the Signal Generator's sample rates read 48000 and a 1 kHz tone 1000, and the Spectral analysis sizes 4096, where some regions' number formats showed 48 000, 1.000 or 4 096. The Signal Generator's fields and the custom target editor's take a point or a comma. A value counts as it is typed, so a click on Generate or Add Target uses it, and a frequency or a count typed as 1,000 or 1.000, one or one thousand depending on the region, is refused with the reason. The raw import sheet gives sizes in English, such as 3.8 MB, and the per-channel statistics name their units and give the DC offset in percent of full scale. File sizes read as Finder shows them, in the sidebar and the report, and the report's date is in English.
- FixTarget levels print as set. A custom target's levels print as entered, to two decimals, in the report's target badges and tooltips: a -1.5 dBTP ceiling reads -1.5, where the badge read -1 and the tooltip -2. The Normalize and Limit True Peak panels opened from a target fill in its ceiling as set, where a -1.25 dBTP ceiling filled in -1.2, and the edit's confirmation names the levels applied. A platform's tooltip reads "Spotify will turn this down by 2.6 dB".
- FixA reading next to its limit shows which side it is on. Where a reading prints beside the limit that judges it, on the delivery card, in the tooltips, on the report's badges and cards, in the sidebar and in the per-channel statistics, it prints as many decimals as it takes to land on the side of the limit its verdict does: a true peak of -0.96 dBTP under a -1 dBTP ceiling reads "-0.96 dBTP is over its -1 dBTP ceiling", where it read -1.0.
- FixThe report's cards and loudness grid fit. The PDF and HTML report's metric cards keep each value and its unit on one line, four to a row, and the loudness curve's grid labels no longer overlap: the grid draws the primary target's reference with its name, such as Apple Music -16 LUFS, and a fixed line close to it gives way.
- FixLabels keep their text. The Output panel's built-in routing presets show their full names, and chapter boundary numbers from 10 up stay on one line.
- FixWindows open without a focus ring. With keyboard navigation on in System Settings › Keyboard, a window, sheet or popover opened with its first control focused, such as the Output panel's close button or the first choice in a picker. Nothing is focused when one opens now, and Tab moves into its controls.
- FixNothing reads -0. A value just under zero prints as 0, such as the raw import sheet's plausibility, which could read -0.00.
- FixLoudness thresholds flag the whole file. The M, ST and TP thresholds in the loudness section flagged only what had played since the file opened, or the range analysed with ⌘Return, so a threshold set on a file just opened showed nothing until it played. They now flag every stretch of the whole file as soon as it has been measured, in the waveform, in the section's counts and in the report.
- FixThe wand selects at the level clicked. Spectral's wand took every connected region louder than the clicked level, so a click between two harmonics of speech could select the whole spectrum for over two seconds. It now takes the connected region within its tolerance above or below the clicked level, reaching 250 ms each side by default. Pointing with the wand outlines what a click would take, and a click replaces the selection: Shift adds and Option subtracts.
- FixSpectral edits apply in full across the selection. The feather's roll-off was centred on the selection's edge, so the edge got half of an edit, the edit reached past the selection where nothing showed it, and a thin selection got less still: a one-bin line took a ninth of the attenuation set. Every selected cell now gets the full edit, and the roll-off lies outside the selection, where the canvas draws it. Solo plays a thin selection, which it silenced.
- FixCompare switches between sample rates without a pause. Switching to a file at another sample rate, such as a codec stream at 44.1 kHz beside a 48 kHz master, paused playback for a moment, and the residual and Auto-align were unavailable for that file. A file at another rate now plays at slot A's rate, resampled once when it is added, so the switch is instant and the residual and Auto-align work. Its readings and reports describe the file itself, at its own rate.
- Fix
specula compare compares files at different sample rates. It refused them. It now resamples the second file to the first file's rate, as Compare in the app does.
- FixA second Specula leaves the first one's renders on disk. Starting a second copy of Specula deleted the WAV files the first copy's Render to Compare Slot had written, while its Compare slots still named them. Each copy now keeps its renders in a folder of its own.
- FixSelecting a speech region leaves Undo alone. A click that selected a region recorded an undo step, so the next ⌘Z changed nothing and Redo was lost. Only a change to a region is a step now.
- FixOpening a file clears the time selection. The previous file's IN and OUT carried over to the next file opened.
- FixThe chapter map stays readable with many chapters. With many chapters in view, each chapter's readings drew across its neighbours'. A chapter now shows what its width holds, from its number alone to every reading, and its tooltip lists all of them. A chapter that fails ACX or stands out from the median shows red at any width.