Voice & Vocals

Extract an acapella with Stem Separation (12.3) and prep it

Any audio clip — your own old track, a field recording of someone singing, a bounce you lost the stems for — can be split into Vocals, Drums, Bass and Other. The vocal stem is the raw material for everything below.

1. Right-click the audio clip → separate into stems (12.3 Suite). Live creates new tracks per stem. Keep the Vocals track, mute the rest (or keep Other for the pad bed).
2. Gate first: separation leaves faint bleed in the silences. Threshold around −40 dB, Attack 1 ms, Release 150–250 ms, Floor −inf; use Lookahead so consonants aren't clipped.
3. HP at 100 Hz (male) / 150 Hz (female), LP at 12 kHz: separation artefacts live at the extremes.
4. Compressor 4:1, 10 ms attack, 100 ms release, 4–6 dB gain reduction. Now the level is predictable for the slicing and pitch devices downstream.
5. Consolidate (Cmd+J) the cleaned vocal to one clip. This is your acapella; save it to the User Library.

Vocal chops: Simpler Slice mode, Sampler and Drum Rack slicing

The chopped-vocal hook: a phrase cut into syllables and re-played as a melody. Three routes depending on how much you want to re-arrange.

1. Drag the acapella into Simpler. Mode: Slice. Slice by Transient at Sensitivity ~50 %, or by Beat at 1/8 for a grid-aligned chop that ignores where the words are (the "stuttery" garage feel). Playback: Mono or Thru; Fade 3–8 ms removes clicks.
2. Each slice is now on a MIDI note from C1 upward. Play the slices on Push; chromatic transposition is not available in Slice mode, so for re-pitched chops use step 4.
3. Alternatively right-click the clip → Slice to New MIDI Track → Drum Rack. Each slice lands on a pad with its own Simpler; per-pad you can add Transpose, filter, Drum Buss. Best for drum-like chops.
4. For melodic chops: slice to a Drum Rack, pick the best 3–4 syllables, drag each pad's Simpler into a new Instrument Rack chain in Classic mode with Warp on (Complex Pro, Formants 100 %). Now each syllable is playable chromatically with preserved formants; or use the MIDI Pitch tool / a Pitch MIDI effect to re-pitch slices in Slice mode per chain.
5. Sampler route: load the acapella, use the sample's loop points per zone and the 3 LFOs for vibrato on long syllables; Sampler's Pitch/Osc page adds FM for an electric edge.
6. HP 180 Hz, +3 dB at 3 kHz for presence. Dotted-1/8 Echo at 25 % and a short room glue a chop melody into a part.

Variations: see the vocal chop rack. Future-bass style: pitch chops +12, Shifter Formant on, massive Shimmer reverb.

Formant shifting: Shifter, Auto Shift (12.1) and Pitch Hack

Changing the apparent size of the singer without changing the note, or the note without the size. Three devices, three characters.

1. Auto Shift (12.1) is the correction/harmonizer: set the Scale to the song key, Amount 80–100 % (100 % = hard-tuned T-Pain effect), Speed fast for robotic snapping or slow for natural. The Formant control shifts the vocal size independently; Harmonize adds voices at intervals (a fifth or an octave); the built-in LFO on pitch gives vibrato or wobble; the external sidechain can tune the vocal to a MIDI/audio reference.
2. Shifter with Formant mode on: Pitch 0 and Formant −5 st = bigger singer, same note. Formant +7 = child/robot. Shifter also has a Delay mode and a Ring mode (ring modulation for metallic vocals); Mix 100 % for a replacement, 50 % for a doubled chorus.
3. Pitch Hack is the glitch route: Interval up/down with Rate, Reverse and Recycle. −12 st with Recycle 40 % gives a descending echo of octaves; Reverse on with a short delay makes backwards-talking artefacts.
4. Always HP afterwards: formant shifts downward push junk below 150 Hz.

Vocoder: carriers, bands and the robot/talk-box

Vocoder imposes the vocal's spectral envelope on a carrier. The carrier choice is the whole sound: Noise for whisper/robot, a synth from another track for the classic talking-synth, Modulator (self) for a strange enhancer.

1. Vocoder on the vocal track. Carrier: External, pick a synth track (Analog, two saws, slight detune, 4-note chords) as the source via the chooser; play the chords live or from a MIDI clip. The chords should be sustained — gaps in the carrier mean gaps in the vocal.
2. Bands: 24–40 for intelligible, 8–12 for the chunky 70s robot. Range 100 Hz–8 kHz. Release 30–80 ms; longer smears consonants into pads.
3. Unvoiced (the noise injected on consonants) 15–30 % keeps the words readable. Formant knob shifts the analysis bands: +3 for a smaller "voice", −3 for a bigger one.
4. Carrier Noise with a narrow Range: the whispering-robot/texture vocoder. Carrier Modulator: the vocal filters itself — a resonant enhancer; set Depth high and it starts to ring like a talk box.
5. Talk-box feel: carrier = a saw synth with a Pedal (overdrive) before the vocoder on the carrier track, 16 bands, Release short, then Amp/Cabinet after the vocoder for the guitar-amp halo.
6. Chorus-Ensemble after for width; vocoders are inherently mono-ish.

Granular vocals: Granulator III and Grain Delay

A single word frozen into a cloud, scanned slowly, or a spoken phrase smeared into a pad. The voice stays recognisable even with ten-millisecond grains.

1. Granulator III with a 1–3 s vocal phrase. Grain size 40–80 ms keeps formants readable; under 20 ms it becomes a buzzy texture. Spray 20–40 % randomises grain position for smoothness.
2. Scan slowly through the file (0.05–0.2 Hz) or freeze the position on a vowel; hold chords on the keyboard — each note is the vowel at that pitch, so a held triad becomes a vocal pad.
3. Add 5–15 % FM for a metallic edge; use MPE pressure → grain size on Push for performance.
4. Grain Delay after: Pitch +12 with Spray 40 ms and 50 % feedback stacks an octave-up smear that shimmers over the pad. Pitch −12 gives the sub-vocal drone.
5. Tides reverb (Hybrid Reverb) with modulation fits granular pads well; keep it 30–40 % wet.

Reverb and delay throws (Echo, Hybrid Reverb) via sends

The last word of a line thrown into a long delay or a huge reverb while the rest stays dry. A send with automation, not an insert.

1. Return A: Echo, dotted-1/4 L and 1/4 R (or Ping Pong), Feedback 55–65 %, feedback filter HP 300 Hz LP 5 kHz, Echo's Reverb section 20 % for softness. Return B: Hybrid Reverb Dark Hall, Decay 6–8 s, Pre-delay 60 ms so the dry word is still clear.
2. Automate the vocal's send to A: −inf all the time, 0 dB for the duration of the last word only (draw a square in the automation lane). The word is thrown; nothing else is.
3. On the return, put a Compressor with sidechain from the dry vocal: 6 dB of gain reduction while the singer sings, releasing in 300 ms — the throw blooms in the gaps and stays out of the way of the words.
4. Echo's Gate/Duck section does the same ducking inside the device if you prefer one device.
5. For a "throw into the next section": throw the last word, then automate the Echo feedback to 100 % for one bar and cut — it doubles as a riser (see FX & risers).

De-essing with Multiband Dynamics; doubling and width

The two mix jobs every vocal needs: tame the 5–9 kHz sibilance so the effects don't spit, and widen without losing the mono centre.

1. Multiband Dynamics: solo the High band, set its crossover so the band covers ~5–9 kHz. In the Above section set threshold so only "s" and "t" trigger (−20 dB region), ratio 4:1, attack 0.5 ms, release 40 ms. 3–6 dB of reduction on sibilants, none on vowels.
2. Do the de-ess before saturation, distortion and reverb. Sibilance through Saturator or a bright Shimmer is the most common ugly vocal artefact.
3. Doubling: duplicate the vocal to two tracks, pan hard L/R, each with Shifter Delay mode: 12–20 ms delay, Pitch +8 cents on one and −8 cents on the other, Mix 100 %. Keep these 6–10 dB below the centre vocal. The centre stays mono and in phase; the doubles give width.
4. Alternative in one device: Chorus-Ensemble in Ensemble mode, Amount 15–25 %, Width 100 %, Dry/Wet 30 %.
5. Check mono: Utility Mono on the master, the vocal must not drop more than ~1 dB. If it does, the doubles are too loud or the delay too short.

Rhythmic gating: Auto Filter (12.2) LFO, Gate sidechain, Beat Repeat

Chopping a sustained vocal or vocal pad into a pattern without slicing — three approaches from smooth to brutal.

1. Auto Filter (12.2 redesign): LP, LFO shape Square (or the Sample-and-Hold shape for random steps), Rate synced 1/8, Amount 100 %, Phase for stereo offset. The filter opens and closes on the grid: the trance "gated vocal pad".
2. Gate with external sidechain from a MIDI track playing a hat pattern (any short sound, muted via the sidechain's own listen — set the sidechain source track's output to Sends Only): Attack 1 ms, Hold 20 ms, Release 60 ms. The pattern you program on the hat track is the gate pattern. Change the clip, change the chop.
3. Beat Repeat with Chance 20–30 % and Variation 3 for occasional stutters that land on the grid.
4. A 1/8 Echo after any of these fills the gaps with tails and keeps the gate from sounding dry.

Spoken word and trip-hop treatments

The dusty, close-mic'd, slightly broken spoken vocal: narrow bandwidth, tape and vinyl artefacts, and a delay that never fully resolves.

1. Compress hard: 8:1, 1 ms attack, 80 ms release, 8–10 dB reduction. Spoken word should be flat and close; the pumping is the aesthetic.
2. Band-limit: HP 250 Hz, LP 4 kHz, maybe a +4 dB bell at 1.5 kHz for the "radio" peak. Automate the LP open for one phrase to make it "lean in".
3. Saturator Analog Clip 6–10 dB for tape density; Vinyl Distortion for crackle and tracing noise; Redux lightly (12 bit, 2× down) for the sampler-era grit.
4. Echo with Wobble and Noise on, feedback 45 %, dark filter — the dub/trip-hop quarter-note tail.
5. Pitch down 1–2 semitones (clip transpose, Complex Pro, Formants 100 %) for weight; or −5 to −7 st with formants not preserved for the hooded, slowed "chopped and screwed" voice.

Variations: For phonk/drill, pitch −6 st, Formant −3 on Shifter, Redux 8-bit. For boom bap, skip Redux, add Drum Buss Drive 20 % on the vocal bus.

Corpus, Resonators and Spectral Resonator: vocals tuned to the key

Instead of tuning the pitch, impose a tuned resonance: the vocal becomes a chord, a bell, a string. Spectral Resonator in particular turns spoken word into harmonic pads with the words still faintly inside.

1. Spectral Resonator: set it to MIDI mode and feed it chords from a MIDI track (sidechain MIDI input in the device). The vocal is spectrally resonated at the chord's pitches. Decay 1–3 s, Harmonics up for a brighter overtone series, Unison for width, Frequency Shift for inharmonic dissonance.
2. Resonators: 5 tuned comb resonators; Note to the key root, intervals of fifths/octaves, Decay 30–50 %, Const resonance so all 5 ring equally. Dry/Wet 50 % — the vocal now has a drone chord underneath every syllable.
3. Corpus: Resonance type String or Membrane, tuned by MIDI (sidechain) or Tune to the root; Decay 50–70 %. Corpus on a spoken vocal makes a plucked-string texture out of consonants.
4. Prism reverb for a crystalline tail; HP everything at 150 Hz because resonators pile up low partials.

Vocal as bass: Auto Shift down two octaves, Operator FM, sidechain

The vowel formants of a vocal are a filter nobody else has. Pitch a vocal down to the bass register, or use it to FM a bass, and you get a growl with a human mouth.

1. Auto Shift: hard tune (Amount 100 %, fast) to the key, then shift Pitch −24 semitones (two octaves). Formant −3 so it doesn't sound like a tiny creature in a large body. A sustained "aaah" at D2 becomes a D0 growl with the vowel shape intact.
2. Saturator Medium Curve 10–14 dB: the bass needs harmonics in the 100–500 Hz range to be audible; the vocal's formants shape where those harmonics peak, which is the "mouth" sound.
3. Auto Filter LP with Envelope Follower amount ~30 % keeps the top end dynamic; the 12.2 Auto Filter's envelope section does this inside the device.
4. Sidechain to the kick, mono below 150 Hz, layer a plain Operator sine sub one octave below if the Auto Shift output is thin in the sub.
5. FM variant: Operator can't be FM'd by audio directly, but Sampler can — load the vowel in Sampler's Osc page as the FM source (Osc → Sample FM) and play bass notes. Or route the vocal through Roar with the Feedback section and a Roar LP so the vocal's dynamics drive Roar's envelope follower into the shaper's drive.
6. Resample the result; the growl is best used as a sample in Simpler so it's playable chromatically with Complex Pro warping.