Repository navigation
Releases
Cantus 0.1.0-beta.16
Cantus 0.1.0-beta.16 — macOS is back, and a song never hears the last one. A free,
open-source digital audio workstation for Windows, Linux and macOS (preview), made by Luis
Ángel Almazán (lalmazan.com).
"Whatever you do, work heartily, as for the Lord" — Colossians 3:23.
Español abajo.
A short release that finishes what beta.15 left: the macOS preview returns, and the one
known issue beta.15 shipped with is fixed. Everything beta.15 brought is here too: seventeen
new effects, two controllers, Chorda and Tympanum.
What is new
- macOS, again: the universal disk image for Apple silicon and Intel is back. It was
held out of beta.15 because one of Fragor's presets went over its CPU budget on GitHub's
Mac; that preset now costs about a fifth less (below). - Fragor costs less, and sounds the same to the bit: the pedals' circuits are solved
about 21 % faster. The heaviest preset, Synth fuzz (the germanium fuzz into the triangle
muff, in stereo), went from 0.35–0.38 of a core to 0.28; every output sample is
unchanged. - The AppImage's new name:
Cantus-0.1.0-beta.16-x86_64.AppImage, as the AppImage
catalogue names programs. The old names are published too, so links and copies that
already exist keep working, and beta.15 finds this update.
Fixed
- An effect stayed silent after you opened another song in the same session (beta.15's
known issue: Aurora's master EQ after Neon Tide). Every song's effects, instruments and
hosted plug-ins now start fresh when another project is opened, made new, loaded from a
demo or recovered: nothing of the old song is heard, and its echoes never reach the new
one. Checked by opening every demo after every other (42 pairs), and by Replace project
with undo and redo. - The plug-in windows, A/B comparisons, selections, Last tweaked, the clipboard, MIDI
mapping and the session view's performance are closed or cleared too, rather than pointing
at the old song's objects.
Known limits
- macOS is a preview: built and tested on GitHub's Mac machines, not yet opened on a
real Mac, and not signed (below). - A Fragor board with two of its heaviest pedals in stereo can use about a third of a core;
the guitar presets (one pedal, mono) use a twentieth to a tenth.
Install
| System | File | |
|---|---|---|
| Windows 11, x64 | Cantus-0.1.0-beta.16-windows-x64-setup.exe |
Run it; no administrator rights. Also a portable zip. |
| macOS 11 or newer, Apple silicon and Intel — preview | Cantus-0.1.0-beta.16-macos-universal.dmg |
See below. |
| Linux x86-64 (Ubuntu 22.04 or newer, or alike) | Cantus-0.1.0-beta.16-x86_64.AppImage |
chmod +x it and open it; nothing to install. |
Each file is also published under a name without the version (Cantus-windows-x64-setup.exe,
Cantus-macos-universal.dmg, Cantus-x86_64.AppImage), and the AppImage also under its old
names (Cantus-0.1.0-beta.16-linux-x86_64.AppImage, Cantus-linux-x86_64.AppImage), the
same file. Check a download against SHA256SUMS.txt if you like. The Windows installer is not
code-signed yet: choose More info ▸ Run anyway.
macOS is a preview. The disk image is universal, built and tested on GitHub's macOS
machines, but it has not yet been opened on a real Mac and it is not signed: the first
time, macOS blocks it — open System Settings ▸ Privacy & Security and choose Open Anyway.
Please tell us how it went, good or bad.
Updating: Cantus tells you this version is out. Songs saved by beta.16 open in beta.15
too, and every older song opens in beta.16.
Start with the quick-start manual.
Feedback
Open an issue or write in
Discussions — say what you did, what
you expected and what happened, and your system. We especially want to hear from Mac users.
Cantus 0.1.0-beta.16 — macOS vuelve, y una canción nunca oye a la anterior
Una estación de audio digital libre y de código abierto para Windows, Linux y macOS (vista
previa), hecha por Luis Ángel Almazán (lalmazan.com).
«Y todo lo que hagáis, hacedlo de ánimo, como al Señor» — Colosenses 3:23.
Una versión corta que termina lo que dejó la beta.15: vuelve la vista previa para macOS,
y se corrige el único problema conocido con el que salió la beta.15. Todo lo que trajo la
beta.15 sigue aquí: diecisiete efectos nuevos, dos controladores, Chorda y Tympanum.
Novedades
- macOS, de nuevo: vuelve la imagen de disco universal para Apple silicon e Intel. Quedó
fuera de la beta.15 porque uno de los presets de Fragor pasó de su presupuesto de CPU en la
Mac de GitHub; ese preset ahora cuesta cerca de una quinta parte menos (abajo). - Fragor cuesta menos y suena igual, bit a bit: los circuitos de los pedales se resuelven
cerca de un 21 % más rápido. El preset más pesado, Fuzz de sintetizador (el fuzz de
germanio hacia el muff triangular, en estéreo), pasó de 0.35–0.38 de un núcleo a 0.28; ni
una muestra de la salida cambió. - El nuevo nombre del AppImage:
Cantus-0.1.0-beta.16-x86_64.AppImage, como nombra los
programas el catálogo de AppImage. Los nombres anteriores también se publican, así que los
enlaces y copias que ya existen siguen funcionando, y la beta.15 encuentra esta
actualización.
Corregido
- Un efecto se quedaba mudo después de abrir otra canción en la misma sesión (el
problema conocido de la beta.15: el EQ del master de Aurora después de Neon Tide). Ahora los
efectos, instrumentos y plug-ins de cada canción empiezan de cero al abrir otro proyecto,
crear uno nuevo, cargar una demo o recuperar uno: no se oye nada de la canción anterior, y
sus ecos nunca llegan a la nueva. Comprobado abriendo cada demo después de cada otra (42
pares), y con Reemplazar proyecto deshaciendo y rehaciendo. - Las ventanas de los plug-ins, las comparaciones A/B, las selecciones, Lo último tocado,
el portapapeles, el mapeo MIDI y la interpretación de la vista de sesión también se cierran o
se vacían, en vez de apuntar a lo de la canción anterior.
Límites conocidos
- macOS es una vista previa: compilada y probada en las máquinas Mac de GitHub, todavía
no abierta en una Mac real y sin firmar (abajo). - Un tablero de Fragor con dos de sus pedales más pesados en estéreo puede usar cerca de un
tercio de un núcleo; los presets de guitarra (un pedal, mono) usan de una vigésima a una
décima.
Instalar
| Sistema | Archivo | |
|---|---|---|
| Windows 11, x64 | Cantus-0.1.0-beta.16-windows-x64-setup.exe |
Ejecútalo; sin permisos de administrador. También un zip portátil. |
| macOS 11 o más nuevo, Apple silicon e Intel — vista previa | Cantus-0.1.0-beta.16-macos-universal.dmg |
Mira abajo. |
| Linux x86-64 (Ubuntu 22.04 o más nuevo, o similar) | Cantus-0.1.0-beta.16-x86_64.AppImage |
chmod +x y ábrelo; no hay nada que instalar. |
Cada archivo se publica también con un nombre sin la versión (Cantus-windows-x64-setup.exe,
Cantus-macos-universal.dmg, Cantus-x86_64.AppImage), y el AppImage también con sus nombres
anteriores (Cantus-0.1.0-beta.16-linux-x86_64.AppImage, Cantus-linux-x86_64.AppImage), el
mismo archivo. Si quieres, compara la descarga con SHA256SUMS.txt. El instalador de Windows
todavía no está firmado: elige Más información ▸ Ejecutar de todas formas.
macOS es una vista previa. La imagen es universal, compilada y probada en las máquinas macOS
de GitHub, pero todavía no se ha abierto en una Mac real y no está firmada: la primera
vez macOS la bloquea — abre Ajustes del Sistema ▸ Privacidad y seguridad y elige Abrir de
todos modos. Cuéntanos cómo te fue, bien o
mal.
Actualizar: Cantus te avisa de que salió esta versión. Las canciones guardadas con la
beta.16 abren también en la beta.15, y todas las canciones anteriores abren en la beta.16.
Empieza por el manual.
Comentarios
Abre un issue o escribe en
Discussions — cuenta qué hiciste, qué
esperabas, qué pasó y tu sistema. Queremos oír sobre todo a quien use Mac.
Cantus 0.1.0-beta.15
Cantus 0.1.0-beta.15 — seventeen new effects, a piano and a drummer. A free, open-source
digital audio workstation for Windows and Linux (macOS comes in beta.16), made by Luis Ángel Almazán
(lalmazan.com).
"Whatever you do, work heartily, as for the Lord" — Colossians 3:23.
Español abajo.
The biggest release of sound Cantus has had. Seventeen new effects, from a guitar amp
modelled circuit by circuit to a reference comparator for mastering, and two new
controllers. Chorda, the piano, sounds closer to the Salamander and plays recorded
pianos too. Tympanum, an acoustic drummer, plays modelled and recorded kits with
microphones you can move. Every new effect is also a CLAP and VST3 plug-in.
What is new
- Arca, amp and cabinet: a guitar or bass amp modelled from its circuit (tube preamp
stages, the tone stack from its component values, a push-pull power stage that sags) into
its cabinet. - Fragor, the pedals: up to three pedals in any order, from five modelled circuits (a green
overdrive, an op-amp distortion, a silicon distortion, a germanium fuzz and a triangle muff),
with a noise gate in front. - Tabula, the pedalboard: put Cantus's own effects on a board as pedals, in any order,
each with its own face. - Vinculum, the bus compressor: VCA, opto (matched to the LA-2A's published figures) and
vari-mu, for glue on a bus or the master. - Comparatio, the reference comparator: load the songs you want to sound like and switch
between them and your mix at matched loudness, or blind, to judge honestly. - Purus, real-time repair: de-hum, de-click and de-reverb while the song plays.
- Lenis, the resonance suppressor: tames harsh, ringing frequencies as they happen.
- Cedo, the ducker: lowers one sound under another — the bass under the kick, the music
under a voice — or pumps it to the beat. - Loquax, the vocoder: 8 to 64 bands, a built-in synthesiser or your own carrier, with
the formant and the bands shifted. - Caelum, the shimmer reverb: a large space with octaves and fifths rising in it, and a
Sustain from a short tail to an endless one. - And more: Saltus (pitch shifter and harmonizer), Rota (rotary speaker),
Fractus (glitch sequencer and grain cloud), Lapsus (time shaper), Cribrum (auto
filter and wah), Vetus (lo-fi) and Libra (utility). - Two controllers: Tactus, an X-Y pad that moves any controls you link to it, and
Calculus, a modulator you write as a formula. - Chorda, toward the Salamander: the modelled piano re-voiced in sixteen steps, with its
tuning and the point where the hammer strikes put right. It now also plays recorded
pianos: the Salamander, bitKlavier's nine microphone images, Accurate-Salamander, two
uprights, or your own SFZ, Decent Sampler or SF2. Its own mixer sends each microphone to a
strip of its own. - In tune, key by key: Chorda's modelled piano was retuned string by string, each key's
unison strings centred on its pitch and tuned with the bridge on — every key from A0 to E6
now lands within 0.2 cent of its design (it was up to 18 cents off), and the deepest beats
are within the Salamander's own. - Tympanum, the acoustic drummer: modelled kits with virtual microphones you can move,
six recorded multi-microphone kits, the kit in its room, each microphone to its own strip,
and a Hybrid voice that puts recorded attacks under the model. The full kit with every
microphone costs 7.3 % of a core. - A new default look, Slate Lapis: deep blue for night, its paper twin for day. If you
chose a theme, it stays yours; all 24 themes are still there. - Ctrl+G: the selected tracks go to a new bus in one step, folded in a folder in the
arrangement (asked for by a user). - Several screens: any area in a window of its own, on any screen, remembered for each
set of screens. - Playback and renders that start right mid-song: from any bar, every effect and
instrument takes its settings at once, so a render from bar 33 is the same as the whole
render from there, and live equals offline. - Delay follows the tempo (note values, dotted and triplet) and is now a CLAP and VST3
plug-in too. - Also: insert and delete time across the whole song, a section dragged with its music,
a ripple mode; bands on the mixer's desk for groups and an instrument's microphones; the
instruments' pictures Drawn or Photographic (now also switched inside Chorda and
Tympanum); a pad's own mixer track; Record before or
after FX; twelve dancers in the Slate cast, and a hidden dancer back with View ▸
Stage or Alt+D; each effect's starting preset your choice.
Fixed
- Moving the mouse no longer lags on Windows: with several windows open, every mouse
movement used to repaint each one, so the pointer and clicks fell behind; now the windows
are painted once per frame. - The mixer's polarity and mono buttons are drawn smooth at every size, not pixelated.
- Arca's knob labels read clearly on its enamel in the light themes.
- On Linux, the desktop entry passes validation (for the AppImage catalogue).
- The window that flickered on many screens (issue #1) holds still, with a watchdog, a
log file and a safe-graphics start; menus read whole at 200 %, and preset menus are no
longer transparent. - The fader-cap selector reaches the desk, and dropdowns are no longer see-through.
- The top bar's menus stay under their word and scroll when they are longer than the room
below them, instead of running off the screen. - Instrument editors open whole at every interface scale, also in a window of their own.
- Gravis at the Free phase sounds the same whatever the block size.
- Cantus starts with less memory (an empty song's own data 5.9 → 5.1 MB).
- A Delay on a Tabula board, a board's pedals in a render, Comparatio's switch while
rendering, and an undo after an ARA edit all behave as they should.
Known limits
- An effect may not process after opening another song. If a project is opened after
another one in the same session, one of its effects can stay silent: it lets the sound
through unchanged, shows no input and does not react to its controls (seen on Aurora's
master EQ after Neon Tide). Meanwhile: close Cantus and open the project first; or remove
the effect and insert it again. Fixed in the next version. - The new effects were tested in Cantus and through emulated hosts; not yet played live
through an audio interface by a user, nor opened in another DAW as plug-ins. Please tell
us how they behave in yours. - Several screens were tested in emulation, not yet on many real monitors.
- The recorded pianos and kits download when you choose them (they are large); the modelled
ones need nothing. - Everything listed for beta.14 still holds, except Delay, which now follows the tempo and is
a plug-in.
Install
| System | File | |
|---|---|---|
| Windows 11, x64 | Cantus-0.1.0-beta.15-windows-x64-setup.exe |
Run it; no administrator rights. Also a portable zip. |
| Linux x86-64 (Ubuntu 22.04 or newer, or alike) | Cantus-0.1.0-beta.15-linux-x86_64.AppImage |
chmod +x it and open it; nothing to install. |
Each file is also published under a name without the version (Cantus-windows-x64-setup.exe,
Cantus-linux-x86_64.AppImage), the names the website's buttons
link to. Check a download against SHA256SUMS.txt if you like. The Windows installer is not
code-signed yet: choose More info ▸ Run anyway.
macOS comes in beta.16. The disk image was held back from this version: on GitHub's Mac
one of Fragor's presets used more of a core than its budget allows, and Cantus does not ship
what its tests have not passed. Windows and Linux are complete.
Updating: Cantus tells you this version is out. Songs saved by beta.15 do not open in
beta.14 or older (they can hold the new effects, Tympanum and Chorda's recorded pianos);
every older song opens in beta.15.
Start with the quick-start manual.
Feedback
Open an issue — say what you did, what you
expected and what happened, and your system. We especially want to hear how the new effects
sound on your music, how Chorda and Tympanum compare with what you use, and how Cantus behaves
on several screens.
Cantus 0.1.0-beta.15 — diecisiete efectos nuevos, un piano y un baterista
Una estación de audio digital libre y de código abierto para Windows y Linux (macOS llega en la
beta.16), hecha por Luis Ángel Almazán (lalmazan.com).
«Y todo lo que hagáis, hacedlo de ánimo, como al Señor» — Colosenses 3:23.
La versión con más sonido que ha tenido Cantus. Diecisiete efectos nuevos, desde un
amplificador de guitarra modelado circuito a circuito hasta un comparador de referencias para
masterizar, y dos controladores nuevos. Chorda, el piano, suena más cerca del
Salamander y además toca pianos grabados. Tympanum, un baterista acústico, toca baterías
modeladas y grabadas con micrófonos que puedes mover. Cada efecto nuevo es también un plug-in
CLAP y VST3.
Novedades
- Arca, amplificador y pantalla: un amplificador de guitarra o de bajo modelado a partir de
su circuito (etapas de preamplificador a válvulas, el control de tono con los valores de sus
componentes, una etapa de potencia push-pull que se hunde) hacia su pantalla. - Fragor, los pedales: hasta tres pedales en cualquier orden, de cinco circuitos modelados
(un overdrive verde, una distorsión de amplificador operacional, una distorsión de silicio,
un fuzz de germanio y un muff triangular), con una puerta de ruido delante. - Tabula, la pedalera: pon los efectos de Cantus en una pedalera ...
Content: Cantus packs (round 40)
Sound packs downloaded on demand by Cantus's sound-pack manager (manifest format 4). Not an application release.
- tympanum-hybrid-jazz-1.zip — Tympanum's hybrid attacks for the Jazz Birch kit, fitted from Virtuosity Drums (CC0; Versilian Studios and Karoryfer Samples, played by Austin McMahon).
- tympanum-hybrid-studio-1.zip — Tympanum's hybrid attacks for the Studio Maple kit, fitted from DRSKit 2.1 by DrumGizmo (CC BY 4.0; credit shown in the app).
Cantus 0.1.0-beta.14
Cantus 0.1.0-beta.14 — a new look, drawn by hand. A free, open-source digital audio
workstation for Windows, Linux and macOS (preview), made by Luis Ángel Almazán
(lalmazan.com).
"Whatever you do, work heartily, as for the Lord" — Colossians 3:23.
Español abajo.
The largest change to how Cantus looks since it began. Every button, icon, knob, fader, meter,
menu and window was drawn again by hand, from a blank sheet, for each size it appears at. The
new look is called Slate, and every theme of earlier versions comes back redrawn in it,
with four new ones. This release also brings a convolution reverb, a window of its own for
Delay, a new logo, and an engine that holds much less memory.
What is new
- Slate, the new look. A calm grey-blue workspace: panels set apart by shade rather than by
lines, one blue accent that only means on or selected, and your track colours shown as a
stripe and a tint until you select a track, which then lights across the window. The type is
Barlow Semi Condensed, with figures that do not shift as they count. Keys and fields are flat
wells; nothing shows "−∞" in silence. Slate Paper is its light twin. - Twenty-four themes in Settings ▸ Appearance, each shown as a picture: Slate Lapis
(a night-blue room, the selection in lapis, and the logo's gold on the clock, the tempo, the
rulers' bar numbers and the loop), Graphite and gold and Studio ink, each in a dark
and a light version. The ten themes of earlier versions are back, redrawn in the new
look, each with its own colours: Cantus Night and Paper, Blue Hour and Daylight, Pewter and
Chalk (the earlier Slate and Paper, renamed), Ink & Gold and Parchment, Ultraviolet and
Lagoon. Four are new: Evergreen Night and Paper (forest green and brass), Bordeaux (wine
and rose gold) and Print (black ink on white and one vermilion). And the high-contrast pair.
Every theme was measured easy to read: what carries meaning at 3:1 against its ground and
text at 4.5:1. If you used one of the earlier themes, Cantus opens in it, redrawn, with your
own colour changes kept. Themes you made yourself are untouched. - Icons drawn for each size: the transport, the tools, the windows, the mixer, the browser
and the rest, each drawn for the pixels it sits on. 316 track icons in 17 families —
drums, guitars, keys, voices and more — give each track a picture before its name and on its
mixer strip: right-click a track ▸ Icon…, or click the icon, and search in English or
Spanish. - A new logo, the metal: a gilded C. It is the program's icon, the taskbar's and the
installer's. - The top bar in one row, with the transport, a large clock, the tempo, the meters and the
windows; on a narrow window what does not fit goes to More, laid out with each shortcut
under its key and the CPU and memory as small graphs. - The mixer and the inspector: strips in three widths (Compact, Wide, Large) that read top
to bottom in signal order, each with its real EQ curve and its track's icon; the desk's
controls share the row of tabs. The inspector's sections fold to one line each: Level
(volume, pan, width and a meter, in place of the old pinned strip), Automation with
Show in the arrangement, and Freeze, which says how many effects were rendered and how
big the render is. Where the desk has room, it shows the next empty strips, numbered, as FL
Studio's does: click one to make the strips up to it. - Cables in four styles: Patch, Flow (the default: a light runs along a cable while
sound goes through it), Subway and Jack (plugs and cords that settle when you drop them).
Point at a cable and its whole route to the master lights up. - The arrangement: sections (named spans of bars above the tracks), folders for
tracks, a mode per automation lane, track heights from 24 to 96, and a tool row — Select,
Draw, Slice, Erase, Mute, Zoom — with Follow, which turns the page as the song plays.
The grid menu is in musical order, with triplets (Shift+3). - The piano roll's keyboard looks like a piano now (ivory and ebony at a real piano's
widths, every C named), with the other clips' notes shown faintly behind. - The channel rack: a drum kit's pads fold under its row; Add a channel and a drop zone
at its foot; a melodic channel shows its notes as a small roll. - The browser: search results grouped by pack or folder; a star keeps a sound in
Favourites; the chosen sound's card has its waveform, play, loop and level. The Sound
packs window shows each pack as a cover, with Get, Installed or its progress. - Aula, a convolution reverb: the reverb of a real space, from its impulse response, with
no added latency. Fifteen rooms, halls, chambers, plates and springs made by Cantus, your own
WAV, AIFF or FLAC files (mono, stereo or true stereo), and a free pack of eight real spaces
from the OpenAIR library (CC BY 4.0, credited). Also a CLAP and VST3 plug-in. - Delay has its own window, in the face of a tape echo: each repeat is drawn at its time,
on its side with ping-pong and darker as the tone takes its highs. Drag the echoes across for
the time and up or down for the feedback. At the tempo sets the time from a note value at
the song's tempo, and What is heard says how long the repeats last. A click on its insert
opens the window; the inspector keeps its quick controls. It sounds exactly as before. - Easier to use: a control moved by automation, a modulator, a MIDI controller or a VCA
wears a mark and says what moves it and how to take it back; the help line at the bottom
always says what is under the pointer; no word on screen is cut off. Recent instruments and
effects are one right-click away,Ctrl+Ffinds a preset for the selected track, every area
can be folded or maximised (double-click its tab),Ctrl+F6moves between areas, and every
menu can be walked with the arrow keys. - Fixed: changing the browser's preview level no longer restarts the sound; the
arrangement's headers keep their width after a start in a small window; choosing Chorda no
longer makes the window wait; a new track no longer lands on a bus that other strips feed
(and no longer renames it); a plug-in that brings down a helper it shares with others is
always the one blamed, even on a busy machine.
Measured
Memory. The audio engine no longer reserves at start everything its largest session could
need: its parts are made as the song needs them, still never while it plays. On the same laptop
as beta.13's measurements, with the machine quiet, five starts each (the median):
| Program — what it opens | Window ready | Done working | Memory (private) | At rest (of one core) |
|---|---|---|---|---|
| Cantus 0.1.0-beta.13 — an empty song | 0.74 s | 1.5 s | 237 MB | 1.6 % |
| Cantus 0.1.0-beta.14 — an empty song | 0.76 s | 1.5 s | 140 MB | 1.1 % |
| REAPER 7.85 — an empty project | 1.76 s | 2.0 s | 62 MB | 2.3–3.7 % |
| Cantus 0.1.0-beta.13 — the Beat starter | 0.76 s | 1.6 s | 239 MB | 1.8–2.0 % |
| Cantus 0.1.0-beta.14 — the Beat starter | 0.77 s | 1.5 s | 143 MB | 2.0 % |
| FL Studio 24.1.2 — its template | 2.49 s | 5.8 s | 133 MB | 6.6 % |
beta.14 was measured on 11 November; beta.13, REAPER and FL Studio on 9 November, on the same
machine. Memory is still lost: Cantus holds 41 % less than beta.13 on an empty song (97 MB)
and 96 MB less on the starter, and still more than REAPER (by 78 MB) and FL Studio (by 10 MB).
The window, the time until it is done working and the CPU at rest are beta.13's, within the
noise, and still ahead of both. Most of the memory left is outside Cantus's own data (about
25 MB on an empty song): the graphics driver, the audio drivers and the system's reserves. That
is the next thing we will map.
The look costs no frames. The whole window playing at 1920 × 1080 draws no more than
beta.13's did, and the 64-strip mixer draws half as much. At rest, nothing is redrawn.
Known limits
- The new look was checked in pictures drawn by the program itself, not yet on many real
screens. If anything overlaps, is cut off or looks wrong at your screen's scale, please
tell us with a screenshot. - The interface's size now follows your screen until you choose one (125 % on a 1920-pixel
screen); a size you chose before is kept. - Some instruments and effects keep their own wooden or vintage faces (Anulus, Phasis, Calor,
Organum…): that is their face, not the theme. Settings ▸ Appearance ▸ Units wear their own
faces turns them off. - Aula's own responses are generated, not recorded; tell us how they sound on your music.
- Dragging a section moves the section itself, not the music under it; moving the music with it is planned.
- Delay's time does not follow tempo changes: At the tempo sets it once. Delay is the one
effect not also built as a CLAP and VST3 plug-in. - Everything listed for beta.13 still holds, except the preview level and the arrangement's
headers, which are fixed.
Install
| System | File | |
|---|---|---|
| Windows 11, x64 | Cantus-0.1.0-beta.14-windows-x64-setup.exe |
Run it; no administrator rights. Also a portable zip. |
| Linux x86-64 (Ubuntu 22.04 or newer, or alike) | Cantus-0.1.0-beta.14-linux-x86_64.AppImage |
chmod +x it and open it; nothing to install. |
| macOS 11 or newer, Apple Silicon and Intel — preview | Cantus-0.1.0-beta.14-macos-universal.dmg |
See below. |
Each file is also published under a name without the version (Cantus-windows-x64-setup.exe,
Cantus-linux-x86_64.AppImage, Cantus-macos-universal.dmg), the names the website's buttons
link to. Check a download against SHA256SUMS.txt if you like. The Windows installer is not
co...
Cantus 0.1.0-beta.13
Cantus 0.1.0-beta.13 — quiet when nothing plays. A free, open-source digital audio
workstation for Windows, Linux and macOS (preview), made by Luis Ángel Almazán
(lalmazan.com).
"Whatever you do, work heartily, as for the Lord" — Colossians 3:23.
Español abajo.
A small release about cost: Cantus now does almost nothing when nothing is playing, starts
on its song without a dancer, and Chorda's piano picture follows its controls without
blurring.
What is new
- At rest, Cantus rests. A mixer strip whose effects are fed silence stops running them
until sound returns — every strip, the master chain and the reverb — and what you hear is
the same to the last bit (we test it against a twin that never rests). The meters, the
effect editors and the inspector's meter no longer redraw the window at your screen's
full rate; the threads that wait for MIDI or for a new audio device sleep instead of
checking a thousand times a second. On our test laptop, the song Cantus starts on went
from about 30 % of a CPU core to about 3 % while stopped. - Nothing dances at start. Cantus opens on the same song as before, the second drop of
Neon Tide, without its dancer. The whole demo (File ▸ Demo songs) and your own songs
keep theirs; add one to any song from the Visual shelf. A dancer that is stopped now
holds its pose after ten seconds with nothing happening, and costs nothing until you move
the pointer, press a key or play (it keeps breathing in the companion window, for
streaming). - Chorda's picture no longer pixelates. Its piano is drawn about four times faster, to
the same pixels, so while you turn Width, Lid or Room you see the whole picture, not
a blurred smaller one. Chorda also designs its strings only when they are wanted (a song
with Chorda, or the instrument picker), not every time Cantus starts.
Measured
Started five times each on the same laptop (AMD Ryzen 7 7735HS, 31 GB, Windows 11, nothing else open), watched from outside by the same
script (the method):
| Program — what it opens | Window ready | Done starting | Memory | CPU at rest |
|---|---|---|---|---|
| Cantus 0.1.0-beta.13 — an empty song | 0.74 s | 1.5 s | 237 MB | 1.6 % of a core |
| REAPER 7.85 — an empty project | 1.76 s | 2.0 s | 62 MB | 2.3–3.7 % |
| Cantus 0.1.0-beta.13 — the Beat starter | 0.76 s | 1.6 s | 239 MB | 1.8–2.0 % |
| FL Studio 24.1.2 — its template | 2.49 s | 5.8 s | 133 MB | 6.6 % |
Memory is still lost: Cantus holds more than FL Studio and REAPER, because its audio
engine reserves in advance everything its largest session could need, so that the audio
thread never has to ask for memory while it plays. Making that reservation grow with the
song is planned work.
Known limits
- Everything listed for beta.12 still holds: Sectio's slice points miss some soft hits; a
pad chain with a latent insert is late in the kit's track stem; changing the browser's
preview level restarts the sound; the arrangement's headers can come up narrower in a small
window; the presets were made by measurement and not all tuned by ear yet. - When an audio interface is plugged in or out, Cantus notices from Windows' own signal; if
it misses one, it looks again within ten seconds.
Install
| System | File | |
|---|---|---|
| Windows 11, x64 | Cantus-0.1.0-beta.13-windows-x64-setup.exe |
Run it; no administrator rights. Also a portable zip. |
| Linux x86-64 (Ubuntu 22.04 or newer, or alike) | Cantus-0.1.0-beta.13-linux-x86_64.AppImage |
chmod +x it and open it; nothing to install. |
| macOS 11 or newer, Apple Silicon and Intel — preview | Cantus-0.1.0-beta.13-macos-universal.dmg |
See below. |
Each file is also published under a name without the version (Cantus-windows-x64-setup.exe,
Cantus-linux-x86_64.AppImage, Cantus-macos-universal.dmg), the names the website's buttons
link to. Check a download against SHA256SUMS.txt if you like. The Windows installer is not
code-signed yet: choose More info ▸ Run anyway.
macOS is a preview. The disk image is universal, built and tested on GitHub's macOS
machines, but it has not yet been opened on a real Mac and it is not signed: the first
time, macOS blocks it — open System Settings ▸ Privacy & Security and choose Open Anyway.
Please tell us how it went, good or bad.
Updating: Cantus tells you this version is out. Songs are saved in the same format as
beta.12.
Start with the quick-start manual.
Feedback
Open an issue — say what you did, what you
expected and what happened, and your system. We especially want to hear: whether Cantus
stays light on your machine while it waits, and anything that did not wake up when it
should (a meter, a dancer, a device you plugged in).
Cantus 0.1.0-beta.13 — en silencio cuando nada suena
Una estación de audio digital libre y de código abierto para Windows, Linux y macOS (vista
previa), hecha por Luis Ángel Almazán (lalmazan.com).
«Y todo lo que hagáis, hacedlo de ánimo, como al Señor» — Colosenses 3:23.
Una versión pequeña sobre el costo: Cantus ya casi no hace nada cuando nada suena, abre su
canción sin bailarín, y la imagen del piano de Chorda sigue sus controles sin pixelarse.
Novedades
- En reposo, Cantus descansa. Un canal del mezclador cuyos efectos reciben silencio deja
de procesarlos hasta que vuelve el sonido — cada canal, la cadena del máster y la reverb —
y lo que oyes es idéntico hasta el último bit (lo probamos contra un gemelo que nunca
descansa). Los medidores, los editores de efectos y el medidor del inspector ya no
redibujan la ventana a toda la frecuencia de tu pantalla; los hilos que esperan MIDI o un
dispositivo de audio nuevo duermen en vez de mirar mil veces por segundo. En nuestra laptop
de prueba, la canción con la que arranca Cantus pasó de un 30 % de un núcleo a
un 3 % con el transporte parado. - Nada baila al arrancar. Cantus abre la misma canción de antes, la segunda subida de
Neon Tide, sin su bailarín. La demo completa (Archivo ▸ Canciones de demostración) y
tus canciones conservan el suyo; añade uno a cualquier canción desde el estante Visual.
Un bailarín parado ahora se queda quieto tras diez segundos sin que pase nada, y no cuesta
nada hasta que mueves el puntero, pulsas una tecla o reproduces (en la ventana de compañía
sigue respirando, para transmitir). - La imagen de Chorda ya no se pixela. Su piano se dibuja unas cuatro veces más rápido,
con los mismos píxeles, así que mientras giras Width, Lid o Room ves la imagen
entera, no una más pequeña y borrosa. Chorda además diseña sus cuerdas solo cuando hacen
falta (una canción con Chorda, o el selector de instrumentos), no cada vez que arranca
Cantus.
Medido
Arrancados cinco veces cada uno en la misma laptop (AMD Ryzen 7 7735HS, 31 GB, Windows 11, sin nada más abierto), observados desde fuera
por el mismo script (el método):
| Programa — lo que abre | Ventana lista | Termina de arrancar | Memoria | CPU en reposo |
|---|---|---|---|---|
| Cantus 0.1.0-beta.13 — una canción vacía | 0,74 s | 1,5 s | 237 MB | 1,6 % de un núcleo |
| REAPER 7.85 — un proyecto vacío | 1,76 s | 2,0 s | 62 MB | 2,3–3,7 % |
| Cantus 0.1.0-beta.13 — la plantilla Beat | 0,76 s | 1,6 s | 239 MB | 1,8–2,0 % |
| FL Studio 24.1.2 — su plantilla | 2,49 s | 5,8 s | 133 MB | 6,6 % |
La memoria sigue perdida: Cantus ocupa más que FL Studio y REAPER, porque su motor de
audio reserva por adelantado todo lo que podría necesitar su sesión más grande, para que el
hilo de audio nunca tenga que pedir memoria mientras suena. Hacer que esa reserva crezca con
la canción es trabajo planeado.
Límites conocidos
- Sigue todo lo dicho para la beta.12: los puntos de Sectio pierden algunos golpes suaves;
una cadena de pad con una inserción con latencia llega tarde en el stem de pista del kit;
cambiar el nivel de escucha del explorador vuelve a empezar el sonido; los encabezados del
arreglo pueden salir más estrechos en una ventana pequeña; los presets se hicieron midiendo
y todavía no se han ajustado todos de oído. - Cuando conectas o desconectas una interfaz de audio, Cantus se entera por la propia señal
de Windows; si se le escapa una, vuelve a mirar en diez segundos.
Instalar
| Sistema | Archivo | |
|---|---|---|
| Windows 11, x64 | Cantus-0.1.0-beta.13-windows-x64-setup.exe |
Ejecútalo; sin permisos de administrador. También un zip portátil. |
| Linux x86-64 (Ubuntu 22.04 o más nuevo, o similar) | Cantus-0.1.0-beta.13-linux-x86_64.AppImage |
chmod +x y ábrelo; no hay nada que instalar. |
| macOS 11 o más nuevo, Apple Silicon e Intel — vista previa | Cantus-0.1.0-beta.13-macos-universal.dmg |
Mira abajo. |
macOS es una vista previa. La imagen es universal, compilada y probada en las máquinas macOS
de GitHub, pero todavía no se ha abierto en una Mac real y no está firmada: la primera
vez macOS la bloquea — abre Ajustes del Sistema ▸ Privacidad y seguridad y elige Abrir de
todos modos. Cuéntanos cómo te fue, bien o
mal.
Actualizar: Cantus te avisa de que salió esta versión. Las canciones se guardan en el
mismo formato que la beta.12.
Empieza por el manual.
Comentarios
[Abre un issue](https://github.com/luis314159/cantus...
Cantus 0.1.0-beta.12
Cantus 0.1.0-beta.12 — sounds, instruments and a new look. A free, open-source digital
audio workstation for Windows, Linux and macOS (preview), made by Luis Ángel Almazán
(lalmazan.com).
"Whatever you do, work heartily, as for the Lord" — Colossians 3:23.
Español abajo.
A grand piano voiced against a recorded one, with its microphones, lid and room in your
hands and the piano drawn in its room; a slicer for your loops; a drum machine whose every
pad can have its own channel and its own effects; over five hundred presets you can hear
before you choose — on the website too — and free sound packs that download when you want
them; a new lead dancer; and the whole interface redrawn by hand.
What is new
- Chorda, the modelled grand piano, model 2. Voiced by measuring it note by note and band
by band against a recorded concert grand (the Salamander Grand), then again after a blind
listening test: a soundboard with a body, the hammers' felt and the action's noises, a bass
with weight, a treble that rings without a metallic edge, a wider range from soft to loud.
New controls set where you listen — Mics from over the hammers to the audience,
Width, Lid (closed to open) and the room's Size — and they glide while you play. 24
presets, from Close and Intimate to Hall Grand. Songs made with the first Chorda sound
exactly as before. Still a model, not samples: no download, about a fifth of a core for 64
voices with the pedal down. We are still listening: in our blind test the recorded
piano was preferred to model 2, and model 2 to model 1; this is the second pass on what
that listener heard. Tell us what you hear. - The piano in its room. Chorda's Play view shows the instrument you chose in a
recording room — a concert grand, a grand, a parlour grand or an upright, in black, white,
walnut or rosewood, the lid where the Lid control puts it, the microphones where Mics
puts them, a booth, a studio, a hall or a church — built in 3-D and rendered by Cantus
itself. Each preset has a look that fits its sound; choose another if you like. The picture
never changes the sound. - Sectio, the slicer. Drop a loop on Exemplar (or choose Slice into a new Exemplar
track in the sample browser) and it is cut into slices — at its own markers if it has
them, else at its hits, on a grid, or by hand — one slice per key from C1, each with its
own pitch, level, pan, reverse, choke, envelope and filter. Drag a slice point, add one,
remove one, or slice again with a preview. Dump to piano roll writes a clip that plays
the loop back exactly at its tempo; at another tempo the slices follow the grid. Random,
Stutter and Quantize are one undo step each, and six slicing styles in the library
re-cut your own loop. Measured honestly: on loops made by Cantus, 92 % of the hits get a
slice within 10 ms; on real, hand-annotated drum recordings, 76 % (88 % within 20 ms) —
short of our 95 % aim, and weakest on brushed jazz. Move a point by hand where it misses. - Pulsus III: every pad on its own channel, with its own effects. Split to strips puts
each pad of a kit on its own mixer strip in one click (and Merge back undoes it); the
split kit sounds exactly like the whole one until you move a fader. Each pad can carry its
own chain of effects before the kit's sum, and a chain that is not playing costs
nothing. Drop a folder of samples or an SFZ kit on the pads and it becomes a mapped kit
(names read in English and Spanish, takes as round robins, dynamics as velocity layers).
Sixteen original kits, eight of them new — 808- and 909-class, trap, electro, boom bap,
house, lo-fi, rock, pop, R&B, reggaeton, cinematic, jazz brushes and more — synthesised by
Pulsus, all levelled alike, and a Kit level. Freeze, the recording mode, VCAs, sends, stems and delay compensation all know the
pads' strips. - A library of 511 presets, heard before you choose. Every instrument and effect has
presets for the styles most produced today — trap and hip-hop, pop, house and EDM, R&B,
lo-fi, rock, latin and reggaeton, ambient and film — tagged by style and role (bass, lead,
pad, keys…; vocal, drums, bus, master…), each with a line on what it is for. The new
preset browser filters by style and role, finds a preset by its English name in either
language, and auditions it — an instrument on a phrase, an effect on your own clip —
without touching your project until you load it (one undo step). Every preset is checked
by a test: none clips, none is silent, and switching presets does not jump in level. - Every unit heard on the website. Each instrument and effect on
cantusdaw.com plays excerpts of its presets — rendered by Cantus,
nothing added after. - Free sound packs, on demand. Sound packs… offers Cantus One-Shots (182 sounds, a
note of every instrument preset, by role) and Cantus Loops (86 loops: two-bar
instrument loops at 120 BPM and the grooves of all sixteen kits) — made with Cantus,
released under CC0, downloaded only when you ask (about 57 and 60 MB). - Fuga and Volta. A dancer now starts as Fuga, a masked ninja with a high kick, an
aerial flip and a tornado spin; Volta, a mecha, pops, locks and jumps on its jets. Both
are our own, with every move and mood the family has. Songs you saved keep the dancer they
had. - A new look, drawn by hand. Knobs, faders, lamps and keys drawn as lit objects;
one family of icons; strips tinted by their colour; a larger top bar with a transport
capsule, a big clock and tempo, and nothing folded away at 1920 × 1080; clips that show
their notes and waveforms at full height; a sample browser with tabs, a table of sounds
with their pack's cover, waveform, tempo and length, a card for the sound chosen and a
player; an insert panel beside the mixer (▾ ▸ Insert panel); the song's summary in the
top bar; a piano-roll keyboard shaped like a piano's. Prefer the old mixer fader?
Appearance ▸ Fader ▸ Classic (the default) or Metal. - Measured against the market. A first scorecard compares 97 features across Cantus and
twelve DAWs from their makers' current manuals — what each has, in part or not — with its
method and its sources (docs/scorecard/). It says where Cantus is behind, too. - Agents can use all of it: seven new tools (
slice_sample,dump_slices,
split_kit_outputs,set_pad_chain,make_kit_from_folder,find_presets,
audition_preset) — 94 in all.
Known limits
- Sectio's slice points miss some soft hits (above); brushed and ridden jazz most.
- A pad chain with a latent insert (a linear-phase EQ, a look-ahead limiter) is late by that
latency in the kit's track stem; the strips' stems and the mix are on time. Cantus's
default inserts have no latency. - Changing the browser's preview level restarts the sound that is playing.
- The arrangement's track headers can come up narrower than you left them when Cantus starts
in a small window; drag them wider. - The presets and kits were made by measurement and checked by tests; they have not all been
tuned by ear yet. Some plucked and struck presets sit a little quieter than the rest, to
keep their loudest notes from clipping.
Install
| System | File | |
|---|---|---|
| Windows 11, x64 | Cantus-0.1.0-beta.12-windows-x64-setup.exe |
Run it; no administrator rights. Also a portable zip. |
| Linux x86-64 (Ubuntu 22.04 or newer, or alike) | Cantus-0.1.0-beta.12-linux-x86_64.AppImage |
chmod +x it and open it; nothing to install. |
| macOS 11 or newer, Apple Silicon and Intel — preview | Cantus-0.1.0-beta.12-macos-universal.dmg |
See below. |
Each file is also published under a name without the version (Cantus-windows-x64-setup.exe,
Cantus-linux-x86_64.AppImage, Cantus-macos-universal.dmg), the names the website's buttons
link to. Check a download against SHA256SUMS.txt if you like. The Windows installer is not
code-signed yet: choose More info ▸ Run anyway.
macOS is a preview. The disk image is universal, built and tested on GitHub's macOS
machines, but it has not yet been opened on a real Mac and it is not signed: the first
time, macOS blocks it — open System Settings ▸ Privacy & Security and choose Open Anyway.
Please tell us how it went, good or bad.
Updating: Cantus tells you this version is out. Projects saved by this version open only in
this version or newer. The Cantus packs are listed from this version on.
Start with the quick-start manual
and the pages on presets,
drum kits
and the dancer;
Chorda and Sectio are in the manual's section 10.
Feedback
Open an issue — say what you did, what you
expected and what happened, and your system. We especially want to hear: how Chorda sounds
to you (and which Mics setting you like), a loop Sectio slices badly (attach it if you can),
presets that sound wrong or too quiet, and anything in the new look that is hard to read or
use.
Cantus 0.1.0-beta.12 — sonidos, instrumentos y un aspecto nuevo
Un estudio de grabación digital libre y de código abierto para Windows, Linux y macOS (vista
previa), hecho por Luis Ángel Almazán (lalmazan.com).
«Y todo lo que hagáis, hacedlo de ánimo, como al Señor» — Colosenses 3:23.
Un piano de cola afinado contra uno grabado, con sus micrófono...
Content: Cantus packs (round 37)
Cantus's own sound packs (CC0), downloaded on demand by the app's sound-pack manager (manifest format 4). Not an application release.
Cantus 0.1.0-beta.11
Cantus 0.1.0-beta.11 — the professional studio. A free, open-source digital audio
workstation for Windows, Linux and macOS (preview), made by Luis Ángel Almazán
(lalmazan.com).
"Whatever you do, work heartily, as for the Lord" — Colossians 3:23.
Español abajo.
Record a heavy song at your interface's lowest buffer, ride VCAs, recall the desk, bring your
outboard gear in on time, drive the mixer from a control surface, clean a take in a real audio
editor, and edit clips with ARA 2 plug-ins — and, for the first time, a build for the Mac.
What is new
- The recording mode. Prepare to record renders every channel that feeds the master —
with its effects, sends and automation, before its fader — and then plays only those files;
the tracks you arm stay completely live. The computer nearly rests, so the buffer can go down
without dropouts. Measured on the development machine with 64 tracks, each with EQ,
compressor, delay and reverb: 224–590 % of each block's time live (it cannot play at
all), 2–4 % in the mode, 0.4–0.9 % with the master baked too; one-minute real-time
runs with 64 and 128 tracks had no disk underruns from 32 to 1024 frames. A bar you cannot
miss shows the buffer, the round trip, the CPU before and now and Leave; faders, pans,
mutes and solos keep working for the headphone mix (or lock them); the master can stay live
or be baked too. Leaving gives you back your project exactly as it was — bit for bit. It is
optional and never switches itself on. What it does not do: shorten your interface's own
round trip (the driver and converters set that) — it lets the buffer go lower. - VCA faders, mixer snapshots and scene cues. A VCA moves its channels' level without
touching their faders or their automation; groups can have one. Snapshots capture the desk
(all channels or some, safe channels left alone) and recall it as one undo step, with A/B;
a cue recalls a snapshot at a bar — levels, mutes, solos, polarity, bypasses and EQ —
with a click-free switch. Solo-safe and recall-safe on every strip. - External instruments and effects. Your hardware synth (MIDI out, audio back) or outboard
processor (send and return on your interface's outputs and inputs) becomes part of a track.
An automatic latency ping measures the round trip to the sample and Cantus compensates it,
re-based when you change the buffer. Offline renders never pretend to play your gear: record
its return, as on any console. - Mackie Control and HUI surfaces. Faders with motors that never fight your hand, banks and
extenders, V-Pots, transport, scribble strips and the time display, meters; set up in two
clicks. Tested byte for byte against the protocols' published notes and a built-in virtual
unit — real hardware reports wanted. - The audio editor (Edit in audio editor… on any audio clip): destructive edits with undo
— cut, paste, silence, fades, gain, normalise (peak, RMS or LUFS), invert, reverse, remove DC,
remove silence, declick — joins snapped to zero crossings with short crossfades;
noise reduction by profile; a spectral view where you select a time-and-frequency
region and turn it down; markers, regions and loop points saved with the file. Each edit is
a new file; the original is kept until you choose Remove unused audio. - ARA 2 plug-ins on audio clips. Bind an ARA 2 plug-in to a clip, open its editor from the
clip, and every edit you make in it is an undo step in Cantus; the clip plays the plug-in's
render, with clip gain and fades still yours. ARA plug-ins run inside Cantus's own process
(not isolated) — the menu says so. - Agents can use all of it: 22 new tools (
prepare_to_record,create_vca,
take_snapshot,recall_snapshot,add_external_unit,ping_external_unit,edit_audio,
denoise_clip,bind_ara_pluginand more).
Install
| System | File | |
|---|---|---|
| Windows 11, x64 | Cantus-0.1.0-beta.11-windows-x64-setup.exe |
Run it; no administrator rights. Also a portable zip. |
| Linux x86-64 (Ubuntu 22.04 or newer, or alike) | Cantus-0.1.0-beta.11-linux-x86_64.AppImage |
chmod +x it and open it; nothing to install. |
| macOS 11 or newer, Apple Silicon and Intel — preview | Cantus-0.1.0-beta.11-macos-universal.dmg |
See below. |
Each file is also published under a name without the version (Cantus-windows-x64-setup.exe,
Cantus-linux-x86_64.AppImage, Cantus-macos-universal.dmg), the names the website's buttons
link to. Check a download against SHA256SUMS.txt if you like. The Windows installer is not
code-signed yet: choose More info ▸ Run anyway.
macOS is a preview. The disk image is universal, built and tested on GitHub's macOS
machines, but it has not yet been opened on a real Mac and it is not signed: the first
time, macOS blocks it — open System Settings ▸ Privacy & Security and choose Open Anyway.
Please tell us how it went, good or bad.
Updating: Cantus tells you this version is out. Projects saved by this version open only in
this version or newer.
Start with the quick-start manual
and the pages on the recording mode,
studio mixing,
external gear and surfaces,
the audio editor
and ARA plug-ins.
Feedback
Open an issue — say what you did, what you
expected and what happened, and your system. We especially want to hear: the lowest buffer
your interface holds with and without the recording mode, your hardware synth after the ping,
a Mackie or HUI surface, an ARA plug-in, and anything on the Mac.
Cantus 0.1.0-beta.11 — el estudio profesional
Un estudio de grabación digital libre y de código abierto para Windows, Linux y macOS (vista
previa), hecho por Luis Ángel Almazán (lalmazan.com).
«Y todo lo que hagáis, hacedlo de ánimo, como al Señor» — Colosenses 3:23.
Graba una canción pesada con el búfer más bajo de tu interfaz, mueve VCA, recupera la mesa,
trae tu equipo externo a tiempo, maneja el mezclador desde una superficie, limpia una toma en un
editor de audio de verdad y edita clips con plug-ins ARA 2 — y, por primera vez, una versión
para Mac.
Novedades
- El modo de grabación. Preparar para grabar pre-renderiza cada canal que llega al máster
— con sus efectos, envíos y automatización, antes de su fader — y luego solo reproduce esos
archivos; las pistas que armas siguen completamente en vivo. La computadora casi descansa y
el búfer puede bajar sin cortes. Medido en la máquina de desarrollo con 64 pistas, cada una
con EQ, compresor, delay y reverb: 224–590 % del tiempo de cada bloque en vivo (no puede
sonar), 2–4 % en el modo, 0,4–0,9 % con el máster también pre-renderizado; pasadas de
un minuto en tiempo real con 64 y 128 pistas sin ningún corte de disco de 32 a 1024
muestras. Una barra imposible de pasar por alto muestra el búfer, la ida y vuelta, la CPU
antes y ahora y Salir; faders, paneos, mutes y solos siguen funcionando para la mezcla de
auriculares (o bloquéalos); el máster puede seguir en vivo o pre-renderizarse también. Al
salir recuperas tu proyecto exactamente como estaba, bit a bit. Es opcional y nunca se
activa solo. Lo que no hace: acortar la ida y vuelta propia de tu interfaz (la fijan el
driver y los conversores) — lo que permite es bajar el búfer. - Faders VCA, instantáneas del mezclador y escenas. Un VCA mueve el nivel de sus canales sin
tocar sus faders ni su automatización; los grupos pueden tener uno. Las instantáneas capturan
la mesa (todos los canales o algunos, dejando en paz los protegidos) y la recuperan en un paso
de deshacer, con A/B; una escena recupera una instantánea en un compás — niveles, mutes,
solos, polaridad, bypass y EQ — con un cambio sin clics. - Instrumentos y efectos externos. Tu sintetizador (salida MIDI, audio de vuelta) o tu
procesador externo (envío y retorno por las salidas y entradas de tu interfaz) se vuelven
parte de una pista. Un ping de latencia automático mide el recorrido a la muestra y Cantus
lo compensa, reajustado al cambiar el búfer. - Superficies Mackie Control y HUI. Faders motorizados que nunca pelean con tu mano, bancos
y extensores, V-Pots, transporte, pantallas y el reloj, medidores; en dos clics. Probadas byte
a byte contra las notas publicadas de los protocolos y una unidad virtual integrada — se
agradecen reportes con hardware real. - El editor de audio (Editar en el editor de audio… en cualquier clip): ediciones
destructivas con deshacer, uniones al cruce por cero con fundidos cortos, reducción de ruido
por perfil, vista espectral, quitar clics, cortar silencios, normalizar por pico,
RMS o LUFS; marcadores, regiones y puntos de bucle guardados con el archivo. - Plug-ins ARA 2 en clips de audio: enlaza un plug-in ARA 2 a un clip, abre su editor desde
el clip, y cada edición es un paso de deshacer en Cantus. Los plug-ins ARA corren dentro del
proceso de Cantus (sin aislamiento) — el menú lo dice. - Los agentes pueden usarlo todo: 22 herramientas nuevas.
Instalar
| Sistema | Archivo | |
|---|---|---|
| Windows 11, x64 | Cantus-0.1.0-beta.11-windows-x64-setup.exe |
Ejecútalo; sin permisos de... |
Cantus 0.1.0-beta.10
Cantus 0.1.0-beta.10 — a dancer for your songs. A free, open-source digital audio
workstation for Windows, macOS and Linux, made by Luis Ángel Almazán
(lalmazan.com).
"Whatever you do, work heartily, as for the Lord" — Colossians 3:23.
Español abajo.
A character that dances over your workspace to whatever plays — for the fun of it while you
work, on your stream, or in a clip to post. It makes no sound: every sample of your song stays
the same, bit for bit.
What is new
- Saltator, the dancer. Add a track and choose Visual ▸ Saltator: a character dances
over the workspace to the beat, with no automation at all. Its numbers — move, travel,
speed, energy, lift, facing, spin, size, place, mood and colour — are ordinary automation
and modulator targets, and MIDI learn reaches them; notes from C3 to D#4 choose its
sixteen moves like a drum map, so a clip changes the dance exactly on the bars where you
drew it. Reacts to makes it move with a channel's level or the whole song's. Its window
is a small theatre: the stage, the characters dancing in their tiles, sixteen move pads to
play, and Dance, Body, Look and Ears. - Six characters of our own: Trino the songbird, Miso the cat, Lana the
lamb, Tic the metronome robot, Corchea the eighth note and Sol — every move and
mood, each in its own style, crisp at any size, on light and dark themes. - Bring your own character (Import, beside the characters in the dancer's window): Lottie
(.json,.lottie), Aseprite and TexturePacker sprite sheets, and DragonBones
skeletons. Cantus reads them with its own players, never runs a script in them, bounds
every size before it reads, shows the file's licence first and maps its animations to
our moves by name (you can change the mapping). The character is copied into the project,
so the project carries it. - Over the workspace or on its own: a dancer floats (move it, resize it), docks or
hides — Alt+D shows or hides the dancers. Alt+Shift+D opens the companion, a
small window above everything that keeps dancing while you work — on Windows with true
transparency, soft shadows and all. - Streaming mode (Ctrl+Shift+D, or View ▸ Streaming mode): a window titled Cantus
Stage for OBS or Streamlabs, with a transparent or chroma-key background, a fixed size and
place kept between sessions, and a light visualiser — the master's level and spectrum, bar
and beat, the song's name and tempo. In OBS use Window Capture with the Windows 10 (1903
and up) method to keep the transparency. Measured for an hour under OBS: about 3.4 % of one
processor core; it pauses when hidden or minimised. - Export video (File ▸ Export video…): the song with its dancers and its audio as an
MP4 (H.264 and AAC), a WebM, a transparent MOV or PNG frames for your editor —
square, vertical or wide, 30 or 60 frames a second, on the stage, transparent, on a chroma
key or black, with a title and the visualiser if you like. The preview shows the file's own
frame at the moment you scrub to. Audio and picture are in sync to the sample. A 30-second
vertical MP4 takes a few seconds. - Agents can make it dance:
add_trackwith"kind": "dancer", its moves in the unit
catalogue, andexport_video.
Video on each system
- Windows: MP4 uses the H.264 and AAC encoders Windows already has. Windows N has
none: install Microsoft's Media Feature Pack, or let Cantus download Cisco's OpenH264. - Linux: for MP4, Cantus offers to download Cisco's OpenH264 (Cisco's own binary,
Cisco's licence shown first; it is never shipped with Cantus). The sound is AAC when your
system has FFmpeg or GStreamer, and MP3 otherwise (the dialog says so). WebM, MOV and PNG
frames need nothing.
Install
| System | File | |
|---|---|---|
| Windows 11, x64 | Cantus-0.1.0-beta.10-windows-x64-setup.exe |
Run it; no administrator rights. Also a portable zip. |
| Linux x86-64 (Ubuntu 22.04 or newer, or alike) | Cantus-0.1.0-beta.10-linux-x86_64.AppImage |
chmod +x it and open it; nothing to install. |
Each file is also published under a name without the version (Cantus-windows-x64-setup.exe,
Cantus-linux-x86_64.AppImage), the names the website's buttons link to. Check a download
against SHA256SUMS.txt if you like: Get-FileHash <file> on Windows, sha256sum <file> on
Linux. The Windows installer is not code-signed yet: choose More info ▸ Run anyway.
Updating: Cantus tells you this version is out. On Windows it installs when you close
Cantus; an AppImage replaces itself (the old one is kept beside it, ending in .old, until
the new one starts). Projects saved by this version open only in this version or newer.
macOS comes next, after a session on a real Mac.
Start with the quick-start manual
(also in Help ▸ Manual) and the pages on the dancer,
importing characters,
the stage and streaming
and exporting video.
Feedback
Open an issue — say what you did, what
you expected and what happened, and your system. Tell us how a clip looked once posted, how
the companion did on your stream, and which character you would like next.
Cantus 0.1.0-beta.10 — un bailarín para tus canciones
Un personaje que baila sobre tu espacio de trabajo con lo que suene — por gusto mientras
trabajas, en tu transmisión o en un clip para publicar. No hace ningún sonido: cada muestra de
tu canción queda igual, bit por bit.
Qué hay de nuevo
- Saltator, el bailarín. Agrega una pista y elige Visual ▸ Saltator: un personaje baila
sobre el espacio de trabajo al compás, sin ninguna automatización. Sus números — paso,
desplazamiento, velocidad, energía, salto, orientación, giro, tamaño, lugar, ánimo y color —
son destinos normales de automatización y de moduladores, y el MIDI learn llega a ellos;
las notas de C3 a D#4 eligen sus dieciséis pasos como un mapa de batería, así que un
clip cambia el baile justo en los compases donde lo dibujaste. Reacciona a lo hace
moverse con el nivel de un canal o de toda la canción. Su ventana es un pequeño teatro: el
escenario, los personajes bailando en sus casillas, dieciséis pads de pasos para tocar, y
Baile, Cuerpo, Aspecto y Oído. - Seis personajes propios: Trino el pajarito, Miso la gata, Lana la oveja,
Tic el robot metrónomo, Corchea la nota y Sol — todos los pasos y ánimos, cada
uno con su estilo, nítidos a cualquier tamaño, en temas claros y oscuros. - Trae tu propio personaje (Importar, junto a los personajes en la ventana del bailarín):
Lottie (.json,.lottie), hojas de sprites de Aseprite y TexturePacker, y
esqueletos de DragonBones. Cantus los lee con sus propios reproductores, nunca ejecuta
un script de ellos, limita cada tamaño antes de leer, muestra primero la licencia del
archivo y asigna sus animaciones a nuestros pasos por nombre (puedes cambiar la
asignación). El personaje se copia al proyecto, así que el proyecto lo lleva consigo. - Sobre el espacio de trabajo o en su propia ventana: un bailarín flota (muévelo, cambia
su tamaño), se acopla o se oculta — Alt+D muestra u oculta a los bailarines.
Alt+Mayús+D abre la ventana de compañía, pequeña y encima de todo, que sigue bailando
mientras trabajas — en Windows con transparencia real, sombras suaves incluidas. - Modo para streaming (Ctrl+Mayús+D, o Ver ▸ Modo para streaming): una ventana llamada Cantus
Stage para OBS o Streamlabs, con fondo transparente o de croma, tamaño y lugar fijos que se
recuerdan entre sesiones, y un visualizador ligero — el nivel y el espectro del máster,
compás y pulso, el nombre y el tempo de la canción. En OBS usa Captura de ventana con el
método Windows 10 (1903 y posterior) para conservar la transparencia. Medido durante una
hora en OBS: cerca del 3,4 % de un núcleo; se pausa cuando está oculta o minimizada. - Exportar video (Archivo ▸ Exportar video…): la canción con sus bailarines y su audio
como MP4 (H.264 y AAC), WebM, MOV transparente o cuadros PNG para tu editor —
cuadrado, vertical o ancho, a 30 o 60 cuadros por segundo, sobre el escenario,
transparente, sobre croma o negro, con un título y el visualizador si quieres. La vista
previa muestra el cuadro del propio archivo en el momento al que te muevas. El audio y la
imagen van sincronizados a la muestra. Un MP4 vertical de 30 segundos tarda unos segundos. - Los agentes pueden hacerlo bailar:
add_trackcon"kind": "dancer", sus pasos en el
catálogo de unidades, yexport_video.
El video en cada sistema
- Windows: el MP4 usa los codificadores H.264 y AAC que Windows ya trae. Windows N no
los tiene: instala el Media Feature Pack de Microsoft, o deja que Cantus descargue
OpenH264 de Cisco. - Linux: para el MP4, Cantus ofrece descargar OpenH264 de Cisco (el binario de Cisco,
con su licencia a la vista; nunca viene con Cantus). El sonido es AAC si tu sistema tiene
FFmpeg o GStreamer, y MP3 si no (el diálogo lo dice). WebM, MOV y cuadros PNG no necesitan
nada.
Instalar
Windows 11 (x64): Cantus-0.1.0-beta.10-windows-x64-setup.exe (también un zip portátil).
Linux x86-64: Cantus-0.1.0-beta.10-linux-x86_64.AppImage (chmod +x y ábrelo). El
i...
Cantus 0.1.0-beta.9
Cantus 0.1.0-beta.9 — AI that works with you: agents, stems, audio to MIDI, mastering. A
free, open-source digital audio workstation for Windows, macOS and Linux, made by Luis Ángel
Almazán (lalmazan.com).
"Whatever you do, work heartily, as for the Lord" — Colossians 3:23.
Español abajo.
Everything here runs on your computer. Cantus has no AI service of its own and sends nothing
anywhere; the models it uses are downloaded from their makers only when you ask.
What is new
- Your AI assistant can work in Cantus. Claude Code, Claude Desktop, Cursor, VS Code,
Gemini CLI and LM Studio can drive Cantus through the Model Context Protocol (MCP):
make tracks, write notes and chords, set the mixer, effects and instruments in their own
units, draw automation, set up the session's scenes and clips, export. Every change is
one step in the undo history, named AI: …, and Ctrl+Z takes it back like any of
yours.
Edits ask you first on a card at the bottom right (Allow, Allow for this session,
Deny); removing or replacing your material, undoing and files always ask. A chip in the
top bar shows who is connected; a click opens the AI activity — every call and what
became of it, Undo all of its changes, Disconnect. - 63 tools made for music: an agent writes a part from a chord chart
(write_part), voices chords smoothly (voice_chords), pedals a piano from the chart,
draws automation in the parameter's units (automate), and checks the harmony with a
music linter — notes ringing across a chord change, clashes, doubled notes. It reads
a catalogue of every instrument and effect with its parameters, presets and uses. - Agents can listen.
auditionrenders a range and answers with numbers — loudness per
section, true peak, the spectrum against a genre's curve, clicks — and a WAV; per-track
analysis and a masking matrix say which tracks fight for the same frequencies. After a
change near the end, only the end is rendered again, sounding exactly as the whole song. - The mastering assistant (right-click the master ▸ Mastering assistant…): choose a
platform — Spotify, Apple Music, YouTube, TIDAL, Amazon Music, SoundCloud, CD, Club or
EBU R 128 — and a genre's balance or a reference track you like; it listens and proposes
a chain of Cantus's own effects with every setting and why, compares with A/B at the
same loudness, and Apply puts it on the master in one undo step. Every knob stays
yours. On our demo songs it lands within 0.07 LU of the target, never over the ceiling. - Separate stems (right-click an audio clip or a track ▸ Separate stems…): a song into
drums, bass, vocals and the rest — or six parts with guitar and piano — by Meta's
HT-Demucs, on your computer, Fast or Best. The parts land on their own tracks under a
bus, playing at the original's level, in one undo step; their clips wear a small AI
badge and the project remembers which model made them. Measured on the standard test set
(MUSDB18-HQ): 8.80 dB SDR, the same as the model's paper. - Convert to MIDI (right-click an audio clip ▸ Convert to MIDI…): a sung or played
line (Melody, by Cantus's own pitch tracker), anything with chords (Notes and chords,
by Spotify's Basic Pitch — note for note what Spotify's own program gives), or a drum
loop (Drums: kick, snare, toms, hi-hats and cymbals, by Cantus's own detector, better
with the optional drum model). The notes land on a new track under the clip, where they
were played, in one undo step. - Settings ▸ AI and Settings ▸ Models: one switch turns the whole AI area off (Cantus
is a complete DAW without it); the models are listed with their licences, sizes and where
they come from, and can be removed.
Connect Claude (or another assistant)
External agents are off by default. To let one in:
- Settings ▸ AI ▸ Allow external agents: On. Cantus opens a door on this computer only
(127.0.0.1), protected by a token readable by you alone. - Copy setup for ▸ your client:
- Claude Code: a command — run it in a terminal;
claude mcp listthen shows
cantus connected while Cantus runs. - Claude Desktop: Save the extension (.mcpb)… and open the file, or paste the
copied lines intoclaude_desktop_config.json. - Cursor, VS Code, Gemini CLI, LM Studio: the JSON for their MCP file.
- Claude Code: a command — run it in a terminal;
- Ask your assistant for something small — "add a bass track and write a line on
Am–F–C–G" — and answer the card in Cantus.
Your assistant brings its own model and its own account: Cantus never sees or uses your
Claude key, and what your assistant sends to its company is between you and it.
Models download on demand
No model is in the installer. The first time you separate or convert, Cantus shows the
model's licence, size and origin and downloads it only when you accept: HT-Demucs from Meta
(MIT, about 84 MB), Basic Pitch from Spotify (Apache-2.0, 242 KB), the optional drum model —
Magenta's Onsets and Frames, converted to ONNX — from Hugging Face (Apache-2.0, 6 MB), and
Microsoft's ONNX Runtime to run them (MIT, 11–15 MB, its telemetry turned off). Every file
is checked against its SHA-256. The licences put no limit on the music you make; the music
you separate or convert is still its owner's.
Install
| System | File | |
|---|---|---|
| Windows 11, x64 | Cantus-0.1.0-beta.9-windows-x64-setup.exe |
Run it; no administrator rights. Also a portable zip. |
| Linux x86-64 (Ubuntu 22.04 or newer, or alike) | Cantus-0.1.0-beta.9-linux-x86_64.AppImage |
chmod +x it and open it; nothing to install. |
Each file is also published under a name without the version (Cantus-windows-x64-setup.exe,
Cantus-linux-x86_64.AppImage), the names the website's buttons link to. Check a download
against SHA256SUMS.txt if you like: Get-FileHash <file> on Windows, sha256sum <file> on
Linux. The Windows installer is not code-signed yet: choose More info ▸ Run anyway.
Updating: Cantus tells you this version is out. On Windows it installs when you close
Cantus (an assistant that keeps Cantus's bridge open does not hold the update up); an
AppImage replaces itself (the old one is kept beside it, ending in .old, until the new one
starts). Projects saved by this version open only in this version or newer.
macOS comes next, after a session on a real Mac.
Start with the quick-start manual
(also in Help ▸ Manual): AI agents (MCP) in chapter 11, The mastering assistant in
chapter 7, and the pages on what an agent can do,
separating stems
and converting audio to MIDI.
What Cantus sends and keeps is in PRIVACY.md.
Feedback
Open an issue — say what you did, what
you expected and what happened, and your system. Tell us how your assistant got on in
Cantus, how a separation or a conversion sounded on your music, and how the assistant's
masters compare with yours.
Cantus 0.1.0-beta.9 — IA que trabaja contigo: agentes, stems, audio a MIDI, mastering
Todo esto corre en tu computadora. Cantus no tiene un servicio de IA propio y no envía nada
a ninguna parte; los modelos que usa se descargan de sus autores solo cuando lo pides.
Qué hay de nuevo
- Tu asistente de IA puede trabajar en Cantus. Claude Code, Claude Desktop, Cursor, VS
Code, Gemini CLI y LM Studio pueden manejar Cantus por el Model Context Protocol (MCP):
crear pistas, escribir notas y acordes, ajustar el mezclador, los efectos y los
instrumentos en sus propias unidades, dibujar automatización, preparar las escenas y los
clips de la sesión, exportar. Cada cambio es un paso del historial de deshacer,
llamado IA: …, y Ctrl+Z lo deshace como cualquiera de los tuyos. Las ediciones te
preguntan primero en una tarjeta abajo a la derecha (Permitir, Permitir en esta
sesión, Denegar); quitar o reemplazar tu material, deshacer y los archivos siempre
preguntan. Un indicador en la barra superior muestra quién está conectado; un clic abre
la actividad de IA — cada llamada y lo que pasó con ella, Deshacer todos sus
cambios, Desconectar. - 63 herramientas hechas para la música: un agente escribe una parte a partir de
un cifrado (write_part), enlaza acordes con suavidad (voice_chords), pone el pedal de
un piano según el cifrado, dibuja automatización en las unidades del parámetro
(automate) y revisa la armonía con un corrector musical — notas que suenan a través
de un cambio de acorde, choques, notas duplicadas. Lee un catálogo de cada instrumento y
efecto con sus parámetros, preajustes y usos. - Los agentes pueden escuchar.
auditionrenderiza un rango y responde con números —
sonoridad por sección, pico real, el espectro frente a la curva de un género, clics — y un
WAV; el análisis por pista y una matriz de enmascaramiento dicen qué pistas se pelean por
las mismas frecuencias. Tras un cambio cerca del final, solo se vuelve a renderizar el
final, y suena exactamente como la canción entera. - El asistente de mastering (clic derecho en el máster ▸ Asistente de mastering…):
elige una plataforma — Spotify, Apple Music, YouTube, TIDAL, Amazon Music, SoundCloud,
CD, Club o EBU R 128 — y el balance de un género o una pista de referencia que te guste;
escucha y propone una cadena con los efectos de Cantus, con cada ajuste y su porqué,
compar...