Voice — Advanced & Experimental Techniques

Vocoder deep-dive: every control and what it does to a voice

Live's Vocoder splits the modulator (your voice, on the track where Vocoder sits) into 4–40 bands, follows each band's level, and applies those levels to the same bands of a carrier. Four carrier choices: Noise (internal noise), External (another track via the sidechain chooser), Modulator (the voice filters itself) and Pitch Tracking (an internal oscillator that follows the voice’s pitch).

Robot vocal: Pitch Tracking carrier, Drift saws, Noise carrier

Three robots. Pitch Tracking follows the singer’s melody with an internal oscillator — monophonic, instant, no MIDI. External Drift lets you replace the melody with a programmed monotone or melody. Noise is the whispering machine.

1. Pitch Tracking: set the tracker’s low/high range to the singer (80–400 Hz male, 150–800 Hz female) to stop octave jumps; a little glide smooths wobbly notes. Choose the brighter oscillator shape — a saw/pulse carries more bands. Bands 12, Retro: instant 80s talking machine.
2. Monotone robot: Carrier External from a Drift track: Osc 1 Saw, Osc 2 Saw +7 st (a fifth) at −8 dB, voice mode Mono, Drift 0 % (machines don’t drift), filter fully open. Hold one note (C3) for the whole verse. The vocal turns into a flat, emotionless announcer.
3. Noise robot: Carrier Noise, Bands 16, BW 60 %, Release 15 ms. No pitch at all — a whisper with mechanical edges. Layer it at −10 dB under the Pitch Tracking version for consonant crispness.
4. Redux after at 10–12 bit and gentle downsampling: robot voices sound more "built" with a little aliasing.

Variations: automate Bands from 40 to 8 over a phrase for a "voice turning into machine" transition; or modulate Formant with an LFO at 0.2 Hz for a seasick android.

Vocoded pads and choirs from Wavetable chords

Instead of making a robot, use the voice to animate a pad: long Release, partial Depth, chords that move independently from the singing. The words become a shimmering vowel-shaped movement inside the pad.

1. Wavetable: Osc 1 from the Formant category (already vowel-shaped), Osc 2 Basics Saw for harmonics, Unison Shimmer 4 voices, filter LP open at 6 kHz. Amp env A 400 ms, R 1.5 s.
2. Route the pad track output to Sends Only (silent on its own) and pick it as the Vocoder’s External carrier.
3. Vocoder: Bands 32, BW 140 % for smoothness, Release 300 ms (the pad breathes instead of articulating), Depth 70 % so the chord sustains between words, Unvoiced 10 % (you don’t want words, you want vowels).
4. Whisper or hum long vowels into it, or feed a spoken-word take. Hybrid Reverb Shimmer 35 %, then a slow Auto Pan-Tremolo for drift.

Vocoded drums: the voice as carrier, the break as modulator

Swap the roles. Put Vocoder on the drum track, carrier External = a sustained vocal (a held "aaah", a granular vocal drone, a vocoded pad). The drums now "speak" the vowel; every hit is a burst of voice. Or flip again: voice as modulator, drum loop as carrier = a vocal made of hi-hats.

1. Carrier = a 4-bar sustained vowel. Granulator III frozen on an "ah" with a held chord works perfectly (see the choir recipe below).
2. Vocoder on the drum track: Attack 1 ms (drums are transient), Release 40–80 ms. Gate around −45 dB so the vowel only appears on hits.
3. Dry/Wet 50–60 %: the original kick keeps its weight, the vowel rides on top. Drum Buss after to restore snap.
4. Flipped (talking hats): vocal = modulator, carrier = a 16th hat loop + a noisy snare roll. Bands 16. The words come out as rhythmic noise — Burial-ish ghost speech when mixed low.

Talkbox emulation: Operator + formant filtering + Vocoder Modulator

A real talkbox pipes a synth into the mouth; the mouth is the filter. Two emulations: voice-driven (Vocoder with a buzzy Operator carrier and high band count) and voice-free (automated formant filters on the synth, no singing needed).

1. Operator carrier: Osc A Saw (or Square for a hollow G-funk tone), Osc B sine at ratio 1 modulating A by 20–35 % for growl, Voices 1, Glide 30–60 ms. Play the melody with slides — the glide is what makes a talkbox talk.
2. Pedal Overdrive on the carrier: talkboxes are driven by small, overloaded amps.
3. Vocoder on the vocal: 40 bands, BW 70 %, Release 15–25 ms — talkboxes are articulate, much more than a 16-band vocoder. Speak the words with exaggerated mouth shapes, in rhythm with the melody.
4. Amp (Clean) + Cabinet 1x12 after the Vocoder gives the boxy, mid-forward halo of a horn driver in a tube.
5. Voice-free version: on the Operator track, EQ Eight with two narrow bells (Q 5–8, +12 dB) and the rest of the spectrum cut with an LP at 3.5 kHz. Map bell 1 to F1 and bell 2 to F2 on a Macro each, and draw vowels: "ah" 730/1090 Hz, "ee" 270/2290 Hz, "oo" 300/870 Hz, "oh" 570/840 Hz. Or use Auto Filter in band-pass with high resonance on the F2 position and morph between vowel positions with automation.

Auto Shift (12.1) as a hard-tune: the snapping effect

Hard-tune is not correction, it’s a synth effect: zero retune time, every note a step, the throat flips audible at each transition. Auto Shift does it natively.

1. Put Auto Shift after HP and compression — consistent level = consistent pitch detection.
2. Set the scale: use the Set’s scale (Live 12 scale awareness) or choose root + scale manually. Remove notes the melody shouldn’t touch (e.g. use minor pentatonic instead of natural minor) — fewer targets = more dramatic jumps.
3. Correction amount 100 %, retune speed at its fastest: notes snap instantly. Turn any humanise/smoothing off.
4. Sing between notes and slide on purpose — the effect lives in the transitions.
5. Leave Formant at 0 for realism, or +2/+3 for hyperpop brightness.

Auto Shift harmonizer: parallel chains, scale-aware intervals

One sung line → three or four voices. Build an Audio Effect Rack with one chain per voice; each chain has Auto Shift with its own pitch shift, all quantised to the same scale so the shifted voices land on scale tones.

1. Group the vocal’s Auto Shift into an Audio Effect Rack (Cmd+G), duplicate the chain three times.
2. On each harmony chain shift the pitch by the interval, and keep correction at 100 % so the voice lands on a scale tone. Small negative Formant on upper voices (−1 to −2) stops them sounding like the same person sped up.
3. Pan voices to opposite sides, 7–10 dB below the lead. Add a few ms of Align Delay on one chain (5–15 ms) so the onsets don’t phase-lock.

MIDI-driven pitch: tune the voice from a clip (Auto Shift sidechain, Shifter MIDI)

The voice stops choosing its notes. A MIDI clip on another track decides the pitch; the voice supplies timbre and rhythm. Speak a phrase, play a melody — it sings your melody. Add vibrato and formant moves from Auto Shift’s modulation.

1. Create a MIDI track with no instrument. Draw the melody. On the vocal’s Auto Shift switch to MIDI input and pick that track as the source (or set the MIDI track’s output to the Auto Shift device).
2. Monophonic lines work best; overlapping notes produce legato jumps. Use the MIDI Note Length tool on the clip to force legato if you want glides.
3. Vibrato: Auto Shift’s LFO on pitch, rate 5–6 Hz, depth 15–30 cents. Ramp it in late in long notes (automate depth) — natural vibrato starts after the note settles.
4. Formant: +1 to +3 to make the "singer" smaller, −2 to −4 for weight. Automate it per section independently of pitch.
5. Shifter alternative: Shifter in Pitch mode with MIDI sidechain from the same clip, Glide 50–120 ms. Formants move with the pitch here, so big intervals make chipmunk/giant effects — useful deliberately. Shifter also exposes its own LFO and envelope follower for wobble.

Shifter modes and Pitch Hack on voice: pitch, frequency, ring, cascades

Shifter has three modes; each does something different to a voice. Pitch Hack is the cheap, glitchy, feedback-friendly cousin.

1. Ring mode for monsters: carrier at the song’s root frequency (e.g. A1 = 110 Hz) keeps the result oddly in key. Modulate the frequency with Shifter’s LFO slowly for a moving metal throat.
2. Frequency mode for alien radio: +60 Hz, Wet 50 % keeps intelligibility while the harmonics go inharmonic.
3. Pitch Hack: transpose −12 or +7, Recycle (feedback) 20–50 % = each repeat moves another interval: a word spirals down octaves. Reverse on for backwards fragments; random-interval option for glitch.
4. Use Pitch Hack on a send with a short delay, throw single words into it (see dub throws below).

Sampler vocal synth from one syllable: zones, loops, round robin

Record one sustained syllable ("aah", "ooh", "mmm") and turn it into a playable instrument with sustained loop, natural attack and variation between hits. Sampler doesn’t preserve formants when transposing, so the trick is to sample at several pitches and keep each zone’s transposition within ±3 semitones.

1. Record: sing "aah" for 3–4 s at C2, Eb2, F#2, A2, C3, Eb3… (every minor third across your range) at the same distance and loudness. Or record once and render the extra pitches with Simpler Classic, Warp Complex Pro, Formants 100 % and resample — formant-preserved transposition baked in.
2. Zones: drop all files on Sampler’s Zone editor; set each root key and spread each zone ±1–2 semitones (Key zones). Formant drift is now never more than 2 st.
3. Loop: Sample tab → Sustain Mode Loop (Forward or Back-and-Forth). Loop start after the attack consonant (~300 ms in), length 300–800 ms in the steady vowel. Crossfade 40–100 ms; if it pulses, move the loop end to a zero-crossing area of similar amplitude. Release Mode: continue to the end so the natural tail plays on key-up.
4. Round robin: record 3 takes of each pitch. Assign them as velocity zones (1–42, 43–85, 86–127) with soft crossfades and put a Velocity MIDI effect before Sampler with Random ±12–20: each note picks a different take. For strict alternation use three chains in an Instrument Rack with chain-select zones and a modulated Chain Selector.
5. Modulation: LFO 1 → Pitch 5–10 cents at 5 Hz (vibrato), delayed via its fade-in; Filter env opens LP from 1.5 to 6 kHz over 200 ms ("wah" of a breath onset); Amp env A 60 ms, R 600 ms.

Vocal chops reinvented: Simpler Slice + Recombine and Time Warp

Slice a phrase, play it straight once, then let Live 12’s MIDI Transformations do the re-arranging. Recombine swaps pitches, durations and velocities between existing notes; Time Warp bends timing along a speed curve. You get chop patterns you wouldn’t have played.

1. Slice by Beat 1/16 (grid-based, ignores words — it chops through syllables, which is what you want). Slices map from C1 upward.
2. Make a 1-bar clip playing slices 1–16 in order as straight 16ths. That reproduces the phrase.
3. Select the notes → MIDI Transformations → Recombine: shuffle or rotate the pitch dimension only. Rhythm stays, slice order scrambles — syllables in a new order on the same grid. Try rotation by 3 for a pattern that still sounds related.
4. Delete a few notes, extend some (holding a slice = longer syllable), then apply Time Warp to bar 2 with an accelerating curve: the chops speed up into the next downbeat like a roll.
5. For pitch: duplicate to a second Simpler in Classic mode holding one chop, and play it chromatically (Warp Complex Pro, Formants 100 %) as an answer phrase.

Granular and spectral choirs: Granulator III scans and Spectral Time freeze

Two routes to an endless choir from a single take. Granulator III plays the voice as grains you can scan and pitch with MIDI; Spectral Time freezes the spectrum of any moment and smears it into a sustained chord.

1. Granulator III: load a 5–10 s sung phrase. Set the position to a sustained vowel. Grain size 80–150 ms for a choir (formants readable), 20–40 ms for breathy buzz, 300 ms+ for smeared phrases.
2. Position randomisation (spray/variation) 10–20 %: grains come from slightly different moments = many singers. Scan slowly (or drift the position with an LFO at 0.03 Hz) so the vowel evolves over 30 s.
3. Play 4-note chords: each note re-pitches the grains. Above ±7 st formants shift (it’s grain playback) — keep chords within an octave of the recorded pitch, or record the source in the middle of your chord range.
4. Spectral Time freeze: on a dry vocal track or after the granulator, Freeze in Manual or Retrigger mode synced to 1 bar with a long Fade: each bar captures the current spectrum and holds it. Spoken words become a slowly changing chord-cloud. Add the delay section with small frequency shift for a rising/falling shimmer.
5. Automate the freeze on the last syllable of a phrase only = the word hangs as a choir pad under the next section.

Spectral Resonator, Corpus and Resonators: breath and speech into chords

A whisper or a breath has energy everywhere — perfect excitation for tuned resonators. Spectral Resonator in MIDI mode turns that noise into whatever chord a MIDI clip plays; Corpus and Resonators add physical "body" tuned to the key.

1. Record 8 bars of breath, whispers, "shh", "ff", slow exhales. High-pass at 200 Hz.
2. Spectral Resonator → MIDI mode, choose the chord track as MIDI sidechain. Polyphony at least the number of chord notes. Decay 1–2 s; Harmonics up for brightness; Unison 2–4 for width; Mod set to Chorus or Wander for movement.
3. Resonators (5 combs) after, root on the key, others on fifth/octave, Const on, Decay 30–40 %, wet 30 %: adds a sustained drone floor.
4. Corpus with MIDI frequency from the same chord track (it follows the lowest/last note): Pipe or Tube for a breathy organ, String for a plucked shimmer on consonants.
5. Every breath now "plays" the chord. Exhale = swell; "t" and "k" = plucks.

Whisper layers, breath design and ASMR-intimate proximity

Whispers have no fundamental — they’re formant-shaped noise. A whisper layer doubled under a sung lead makes it feel closer and more secretive. Going further, the "inside your head" ASMR sound is a production decision: close, compressed, bright, dry.

1. Whisper layer: record a whispered double of the lead, timed tightly (edit it to the lead’s consonants). HP 300–500 Hz, air shelf +3–5 dB at 8 kHz, compress hard. Sit it 8–14 dB below the lead; you should notice it when muted, not when playing.
2. Breath design: collect inhales between phrases into a separate track. Use them as rhythmic elements: a pitched-down inhale (−5 st) before a drop is an organic riser. Reverse an exhale for a pre-swell.
3. ASMR proximity: record 5–10 cm from a cardioid mic (proximity effect boosts lows); then shape with EQ: keep 100–200 Hz warmth (+2 dB), no mud cut, gentle 3 kHz dip (−2 dB) to remove harshness, big air shelf at 10 kHz (+4 dB).
4. Compression: two stages, each 3–4 dB — Compressor 3:1 fast for peaks, then Glue Compressor 2:1 slow attack. Mouth sounds, lip clicks and tiny breaths become audible: that IS the effect.
5. No reverb. Any room places the voice at a distance. If you need width, use stereo doubles a few ms apart, not ambience.

Radio, telephone and megaphone voices

Three band-limited voices with different distortion characters. The band is what sells it; the distortion tells the listener what device it’s coming through.

Dub vocal techniques: last-word throws, spring, reverse swells, filtered feedback

The dub engineer’s toolkit on a voice: everything is sends, everything is ridden by hand, and the repeats are more important than the words.

1. Last-word throw: map Send A to a Push encoder or a MIDI controller knob, or automate a square pulse under the final word only. Echo dotted (3/16) on one side, 1/4 on the other, feedback 60–75 %, filter 400 Hz–3 kHz so the repeats get darker and thinner than the dry voice.
2. Spring: Hybrid Reverb in Convolution mode with a spring IR (or any spring impulse you own), 100 % wet on a return, HP the return at 250 Hz. Hit it with single words and percussive consonants for the "boing" — then push Send B on the snare too, so voice and snare share the same tank.
3. Reverse reverb swell: duplicate the phrase, reverse the clip, run it into Hybrid Reverb 100 % wet, long decay, and resample to a new track. Reverse that recording back: the reverb now swells into the first word. Place it so the swell peaks exactly on the dry word’s onset; fade the last 50 ms.
4. Filtered feedback: Send C: Auto Filter band-pass with resonance before an Echo at 85 % feedback. Ride the filter frequency up during the repeats — the throw turns into a sweeping siren. Keep a Limiter at −1 dB at the end of that return.
5. Ducking: Echo’s Ducking section on, so repeats sink while the singer sings and bloom in gaps.

Deeper: dub FX & delays, dub delay send rack, dub reggae.

Dark trip-hop vocal chain: compression, tape, plate, ducked reverb

The intimate, slightly smoky female/male lead floating above slow, heavy beats: close and compressed in front, a dark plate behind it that only breathes in the gaps. 85–95 BPM.

1. Two-stage compression: 4:1 with medium attack (keeps consonants), then Glue Compressor for density. Total 6–8 dB.
2. Tape: Saturator Analog Clip 2–4 dB with Color on, or a parallel chain with Vinyl Distortion crackle at −20 dB for "sampled from vinyl" texture. A tiny pitch wobble via Echo’s Wobble on a 0-feedback, 100 % wet insert (mix 30 %) adds tape flutter.
3. Plate on a send: 2.5–3 s, pre-delay 30–50 ms, LP at 5–6 kHz so the tail is dark.
4. Sidechain the reverb return from the dry vocal: −6 dB while words happen, back up between lines. The voice stays close, yet the space is huge.
5. Automate Send to a 1/4 Echo on line ends only. Don’t widen the lead; it lives in mono dead centre.

Context: trip-hop & downtempo, trip-hop track deconstruction, tape & lo-fi character.

Dubstep / grime MC chain: upfront, dry, aggressive

MCs need to cut through a 140 BPM sub and square-wave leads without reverb wash. Bright, compressed hard, parallel-distorted, with ad-libs and hype doubles processed differently from the main bars.

1. HP at 120 Hz minimum — the sub owns everything below. Cut 200–300 Hz boom, push 3–4 kHz presence.
2. Compressor 8:1, fast, 6–10 dB GR: bars must be at the same loudness from the first syllable to the last breath.
3. Parallel distortion: an Audio Effect Rack with a dry chain and a Roar chain (Serial: Soft Sine into Hard, drive moderate, post-filter LP 7 kHz) at 25–35 %. Adds grit and density without losing words.
4. Hype doubles / ad-libs: separate track, Shifter or Auto Shift Formant −3, pitch −2 st, HP 300 Hz, more distortion, panned ±50 %. Grime ad-libs are usually louder than you’d expect.
5. Short 1/8 Echo at 15 %, Gate/Duck on so it only speaks in gaps. Reverb only on the last word of a 16.

Context: grime, dubstep, grime track deconstruction.

Demonic / pitched-down voice: Shifter −12, formants, layered monsters

A convincing demon is three layers: the original voice for intelligibility, an octave-down layer with formants pushed down even further for size, and a noisy, distorted texture for menace.

1. Octave down with Shifter Pitch −12 (longer window, 50–80 ms, for smoother lows). Then Auto Shift with correction off, Formant −3 to −5: now the throat is huge as well as low.
2. Growl: Shifter Ring mode at 30–50 Hz gives amplitude rattle; Roar with Rectify/Fold for teeth. LP at 1–2 kHz so it’s a rumble under the words, not a hiss.
3. Pitch Hack −12 with Recycle 30–50 %: each word echoes down octaves into the floor.
4. Speak slowly and lower your own voice when recording — any device sounds more demonic on a performance that is already dark. Whispered input gives a "possessed whisper" variant.

Choir from one voice: harmonies, Chorus-Ensemble, doubles, Align Delay

Choir = many voices, slightly different timbres, timing and pitch, spread across space. From one take you can fake all of that; from four quick re-takes you can make it genuinely convincing.

1. Start from the harmonizer rack above and extend to 6 chains: unison, third, fifth, octave up, octave down, unison #2.
2. Per chain: different Formant (−2…+2) = different singers; Align Delay 4, 9, 13, 18, 22, 30 ms = imperfect entries; Auto Shift correction 60–80 % with slow retune so the voices drift slightly in pitch relative to each other.
3. Chorus-Ensemble (Ensemble mode) on the whole rack, Width 100 %: the classic string-machine thickening. Pan chains across the field.
4. Choir EQ: dip 2.5–3.5 kHz by 2–4 dB (stacked presence gets harsh), HP at 150 Hz.
5. Better: record 4 real takes of the same line (unison), comp, then run the rack on each with different chain settings. Real timing variation beats any delay.

Vocal stacking and doubling: ADT, Haas, micro-pitch

Three ways to double a voice, each with a different stereo fingerprint. Choose by how much mono-compatibility and realism you need.

1. Doubles always quieter (−6 to −10 dB) and darker than the lead; HP them higher (200 Hz) so the low-mids don’t thicken in mono.
2. Haas delays must differ between doubles if you use two: equal delays on both sides recreate a centred comb.
3. Check on Utility Mono at the master: lead loss should be under ~1 dB.
4. Stacks for choruses: lead + 2 real doubles + octave up whisper + harmony pair. Bus them, Glue Compressor 2:1, 2 dB GR, one shared reverb.

Beatboxed drums into a Drum Rack kit (Convert Drums to MIDI)

Beatbox a groove into a mic. Live converts it to a MIDI drum pattern (kick C1, snare D1, hat F#1 style mapping) and you replace or layer the sounds — or slice your own mouth sounds into a kit and keep the performance.

1. Record dry, close (10–15 cm), with distinct sounds: "b"/"pf" for kick, "k"/"pch" for snare, "ts"/"t" for hat. Gate and compress before converting.
2. Right-click the clip → Convert Drums to New MIDI Track. Live detects kick/snare/hi-hat-ish hits onto separate notes with a Drum Rack.
3. Fix misdetections by hand; quantise with the MIDI Quantize transformation at 50–70 % to keep human swing.
4. Your mouth as the kit: right-click the audio → Slice to New MIDI Track → Drum Rack (by transient). Keep one best example per sound, assign to pads. Layer each with a synthetic partner: mouth kick + sine kick from Drum Sampler’s Sub, mouth snare + Operator noise snare.
5. Drum Buss on the kit with Boom tuned to the key (e.g. 45 Hz) — gives mouth kicks a real low end.

Vocal as bass: Convert Melody to MIDI → Operator, voice as modulator

Hum or sing a bassline. Live extracts the notes; Operator plays them two octaves down. Then use the original voice to shape the bass — its envelope or its vowels — so the bassline keeps the "talking" of the hum.

1. Hum (closed mouth "mmm" tracks best) or sing "doo-doo" with clear note starts, dry, in tune.
2. Right-click → Convert Melody to New MIDI Track. Clean the clip: delete ghost notes shorter than 1/32, fix octave errors, quantise.
3. Transpose −24 (Pitch MIDI effect or select all + shift) and play Operator: sine A, sine B at ratio 2, FM 20–30 % for harmonics so it’s audible on small speakers. Glide 30 ms for vocal-like slides.
4. Keep the talk: put an Envelope Follower (M4L) on the hum track mapped to Operator’s Osc B level: the hum’s syllables now modulate the bass’s growl. Or a low-band Vocoder (12 bands, Range 60 Hz–2 kHz, Depth 50 %) using the hum as modulator — vowel-shaped bass.

Stem Separation (12.3) on your own archive and Burroughs-style cut-ups

Your old live-set recordings, voice memos, interviews with yourself, rehearsals: Stem Separation pulls the voice out of the room. Then the cut-up method — splice text into fragments and reassemble them by chance — becomes a Session-view game.

1. Right-click the recording → separate stems (12.3 Suite). Keep Vocals. Live recordings with PA bleed separate surprisingly well; MC chatter between tracks, crowd shouts and your own mic talk become usable.
2. Clean: gate (separation leaves "underwater" residue in pauses), band-limit. Accept some artefacts — in a cut-up they read as tape grain.
3. Cut-up, Session view: slice the speech by transient to word/phrase level, Slice to New MIDI Track → Drum Rack. Put 8–16 one-bar clips in Session slots, each triggering different pad orders. Set Follow Actions to "Any" / "Other" with Chance so the clips reorder themselves endlessly — the machine assembles new sentences.
4. Or use the Random MIDI effect (Choices 8, Chance 40 %) in front of the Drum Rack: each word-hit may be replaced by another word.
5. Record the result to Arrangement, keep the accidental phrases that mean something.

Live vocal processing for performance: latency, monitoring, Push throws, feedback safety

A live voice through a laptop in a club: latency must be low, the vocalist must hear themselves, throws must be one press away, and nothing may ever feed back into a howl.

1. Latency: buffer 64–128 samples at 48 kHz (~3–6 ms round trip plus converter latency). Enable Options → Reduced Latency When Monitoring. Avoid lookahead devices (Limiter lookahead, Multiband with long lookahead, spectral devices, Granulator) in the dry path; they add delay.
2. Monitoring: give the vocalist a direct hardware monitor (interface direct mix) for the dry voice, and only the effects come back from Live. Then latency doesn’t affect their timing. Track monitor set to In, but the monitor send carries only wet returns.
3. Throws on Push: map the Throw macro to a Push pad via a momentary mapping (or an M4L button) and Feedback to an encoder. Throw = press on the last word, release. Map the filter macro next to it for dub rides.
4. Feedback safety: HP 120 Hz on the vocal (stage low end feeds back first), a narrow notch at the ringing frequency found at soundcheck (EQ Eight bell −8 dB, Q 8), Echo feedback macro capped at 95 %, and a Limiter at the end of every return. Never put Spectral Time freeze, Pitch Hack recycle or long feedback on a chain that also feeds the stage monitors.
5. Gate the mic (−45 dB) between lines so crowd noise doesn’t trigger the vocoder or pitch tracker.

Clean capture: home recording, de-essing and de-breathing

Heavy processing amplifies sibilance and breaths: a vocoder turns every "s" into a hiss blast, a compressor at 8:1 makes inhales as loud as words. Clean before you create.

1. De-breath by editing, not processing: split clips at each inhale and lower the clip gain 6–12 dB (don’t delete — zero breath sounds robotic). Fade 10 ms either side.
2. De-ess twice in heavy chains: once before compression (so the compressor doesn’t react to "s"), once after saturation/distortion (distortion creates new high-frequency content).
3. Find the sibilant frequency: sweep a narrow EQ Eight bell +12 dB from 4 to 10 kHz; where "s" hurts most is your band centre (male ~5–7 kHz, female ~6–9 kHz).
4. Gate with Floor at −15 dB (not −inf) as a breath/room reducer: quieter between phrases but still natural.
5. Before a vocoder: also reduce plosives (HP 100 Hz + Compressor with fast attack), or every "p" opens all low bands.

Recording at home. Every recipe above is easier on a clean, consistent take, and home voice recording is mostly about distance and headroom.

1. Angle the mic slightly off-axis (15–20°) for plosive reduction without a pop filter.
2. Record a 10-second room-tone take: useful for gating and for noise reduction elsewhere.
3. Record multiple full takes into Arrangement with Take Lanes; comp the best syllables (Live 12 comping), then consolidate.
4. Commit cleanup (HP, de-breath edits) before creative processing — save the clean version as its own clip.