Piano roll
Drag on empty space to select the notes in a box. Shift adds them to the selection, and Shift + click adds or removes a single note.
Dragging a selected note moves all the selected notes. Cmd + drag duplicates them.
Cmd + A selects all notes. Cmd + X, C and V cut, copy and paste. Pasted notes go to the marker, which you set by clicking empty space, and replace the notes under them.
Added Cut, Copy, Paste and Delete to the menu. They work on the selected notes or on the selected text in the Text tab.
The selection stays the same when you switch between the piano roll and the Text tab.
A note named - keeps singing the vowel before it at its own pitch, so held vowels can follow the melody. In the Text tab, -:G4 does the same.
Curves
Curve points snap to the grid and to each curve's steps (semitones for pitch bend). Alt + drag moves a point without snapping.
Voice
Added a Glide slider to the Controls tab, from 0 to 500 ms. It sets how fast the pitch slides between legato notes, for the Classic voices too, and can be automated in the host.
MIDI
Keys on MIDI channel 2 change the sung pitch without changing the words. Roboto sings the held key in place of the written note, one note at a time, and slides between keys at the Glide speed. When you let go, it returns to the previous held key or to the written notes.
Keys on the other channels still start bars from C1 up.
Phrase triggering
MIDI keys from C1 up start their phrase on the exact sample of the note.
While the host plays with Sync on, the phrases after a key stay on the host beat.
Singing starts at the first consonant instead of after a 300 ms lead-in. This also applies to Sing, Cmd + Return and Sing from marker.
Host
Added Follow transport to the Controls tab (plugin only). With Sync on, the whole song plays with the host from its position, with no MIDI keys.
Offline renders no longer leave gaps.
Pronunciation
Syllables written as +na are spelled together with the rest of their word, so a Spanish r inside a word (enamoré, tostadora) is tapped instead of rolled.
Talk box
Added a talk box mode. A plucked string replaces the voice and the mouth shapes it, like a guitar or synth talk box.
The string is plucked at each vowel and goes through an instrument drive, a speaker driver and a tube before the mouth. Consonants stay clean.
Switch it on with the Talk box switch on the Controls tab, or with @talkbox on in the song text. It can also be automated in the host.
It works with every voice, the Classic voices included, and keeps the loudness of the singing voice.
Added the Talk Box Groove built-in song.
Song text
Added @transpose to the Text tab, from -36 to 36 semitones. Like the other @ lines it can set the whole song or change it part way.
Cmd + Return (Ctrl + Enter on Windows and Linux) sings from the note at the cursor. On the Text tab the Sing from marker button starts there too.
The word being sung is highlighted in the text while the song plays.
Controls
While the song plays, the sliders and the Double track and Talk box switches show what is being sung, @ line and curve changes included. They go back to their own values when it stops.
Picking a Classic voice, or going back from one, while the song plays now changes the voice right away.
Interface
The desktop window can be made smaller, down to 770 x 497.
The window size is saved with the plugin state.
The discoDSP logo opens the Roboto page.
iOS: the keyboard no longer covers the text box or the note editor, pinch zooms the roll in time steps, copy and cut follow the selection, the controls scroll at large sizes, panning the roll keeps to one direction, and the scroll bars follow the finger.
Voices
Added Classic voices: Robert, Sarah, Tracy, Andy, Abe, Webster, Richard, Desmond, Male Breath and Female Breath. Pick them from the Classic submenu in the voice box or in Menu > Voices.
Classic voices are sung by a second engine, a formant synthesizer with the sound of 1990s singing software.
The voice sliders and curves shape Classic voices too: transpose, pitch bend, vibrato, breath, consonant level, formant shift, formant bandwidth, volume, double track and reverb.
The Classic voice is saved with the song and can be automated in the host as the Classic voice parameter.
Song text
Added @voice to the Text tab. A line such as @voice Tracy picks a voice preset or a Classic voice by name, and the Text tab writes the current voice the same way.
@ lines can now change the song part way through. Before the first note they set the whole song, further down they apply from the next line on, so @voice, @tempo, @chorus and @language can change between verses.
A voice set by an @ line uses its preset values for that part of the song.
Tempo changes are written to MIDI and karaoke MIDI exports, and LRC and SRT times follow them. When Roboto follows the host tempo, the host tempo is used instead.
Languages
German, French, Spanish, Japanese, Russian, Finnish and Swedish are now sung with the sounds of each language instead of English ones.
Each language has its own vowels, taken from published measurements of its speakers.
Spanish follows the pronunciation of Spain: z and soft c are said like the English th, b, d and g soften after a vowel, and the single r is a firm tap.
P, t and k are said without a puff of air in Spanish, French, Japanese, Russian and Finnish, and b, d and g are voiced all the way through.
German, French, Spanish, Finnish and Swedish use a clear l.
Russian and Finnish have a firmer r.
The Swedish long u has its own vowel.
Export
Added AAC audio (.m4a) to the Export menu: 48 kHz stereo at 256 kbps, with peaks normalized to -0.1 dB.
AAC export is available on macOS, Windows and iOS.
Languages
Roboto now sings in German, French, Spanish, Japanese, Russian, Finnish and Swedish, with spelling rules for each language.
Added a Language menu in the Controls tab. The song text can set it with @language.
Japanese kanji are read through the system dictionary on macOS and iOS.
Added built-in songs in each language.
Voice
New sounds: pure and front rounded vowels, nasal vowels, tapped and rolled r, the ach and ich sounds and held consonants.
Text
Added a Musicalize button next to the tabs in the Text tab. Plain lyric lines get a pentatonic melody with one note per syllable. Lines that already have notes stay as they are.
Export
Added Video (.mp4) to the Export menu: the robot, the discoDSP logo, the waveform and the lyrics as they are shown while singing, at 1280x720 and 60 fps.
The video is H.264 with AAC audio at 256 kbps (192 kbps on Windows). The audio is normalized to a peak of -0.1 dB.
The export runs in the background with a progress bar in place of the waveform and a button to cancel it.
Video export is available on macOS, Windows and iOS.
Fixes
On macOS the voice and language menus no longer open a second time after picking an item.
Import
Added Import to the menu for MIDI, karaoke MIDI (.kar), UltraStar (.txt), LRC lyrics (.lrc), SRT subtitles (.srt) and Roboto song text (.txt) files.
MIDI files bring in the melody of the first track that has notes. Lyric events in the file become the sung words, and notes without lyrics are sung on "ah".
In karaoke files the track that follows the lyrics is used, since the melody is often not on the first track.
UltraStar files keep their pitch, syllables and tempo. Only the first singer of a duet is imported.
LRC and SRT files spread the words of each line over its time on a single note, ready to be given a melody in the piano roll. LRC word times are used when the file has them.
Melodies outside the piano roll range are moved by whole octaves to fit.
Files can be dropped on the window to import them on macOS, Windows and Linux.
Export
Export is now a submenu with WAV and MIDI, plus new karaoke MIDI (.kar), UltraStar (.txt), LRC lyrics with word times (.lrc), SRT subtitles (.srt) and song text (.txt) formats.
Times in the LRC, SRT and UltraStar files match the exported WAV, so they can be used with it directly.
iOS
Import and export work in the standalone app and inside AUv3 hosts, using the Files app.
Initial Release
Singing robot voice synthesizer that turns lyrics and notes into a sung vocal, with English words converted to phonemes and syllables.
Available as VST3, AU and standalone app on macOS, VST3 and standalone on Windows and Linux, and AUv3 and standalone app on iOS.
Editing
Piano roll with a word on each note, snap, zoom and a marker to sing from.
Text tab to write the song as text, with rests, held notes, syllables, tempo and chorus settings.
Curves for volume, pitch bend, vibrato depth, breath, formant shift and chorus along the song.
Words can be given an exact pronunciation with phonemes, for example S-T-R-IY-T.
Voice
Controls for transpose, tempo, vibrato, jitter, breath, consonants, formant shift, formant bandwidth and reverb, all automatable in the host.
Voice presets including Robot, Soprano, Bass and Choir, and a double track for chorus width.
Built-in songs to start from.
MIDI
Keys from C1 up start singing from bar 1, 2, 3 and on while held.
CC 7, the mod wheel, CC 2, CC 74, CC 93 and the pitch wheel control the curves live.
Tempo follows the host, and the standalone app has a MIDI menu for input channel, devices and Bluetooth MIDI on iOS.
Export
Export the song as a WAV file, or as a MIDI file with tempo, notes, words and the curves as controllers.
GUI
Animated robot face, vowel chart with the formant trace, waveform and lyric line while singing.
Interface size from 75% to 200% and a built-in help window.