Experimental, Musique Concrète, Drone & Sound Art

Schaeffer’s closed groove, speed change and reversal: clips as a phonogène

Schaeffer’s first tools were the sillon fermé (a locked groove on a lacquer disc: a fragment looping forever), playback speed change (pitch and time together) and reversal. In Live each maps to one switch.

1. Closed groove: drop a recording into an audio clip, Warp off, loop on, and set the loop brace around a fragment (0.2–2 s). A door creak looped becomes a rhythm, then a timbre — Schaeffer’s discovery that repetition strips a sound of its cause.
2. Speed change: with Warp off, clip Transpose changes speed and pitch together, like tape. −12 st = half speed, octave down, double length. Use Re-Pitch warp mode if you want tape behaviour that also follows the set tempo.
3. Reversal: the clip’s Rev button (or Simpler’s reverse). A reversed piano note is the textbook concrète object: swelling attack, sudden cut.
4. Phonogène: load the fragment into Simpler Classic, Warp off, loop on; play it from a keyboard or Push in chromatic mode — each key is a tape speed. Schaeffer’s phonogène was exactly this: a keyboard of tape speeds.
5. Layer the same fragment at 3–4 speeds (e.g. 0, −7, −12, −19 st) for an instant concrète chord with natural speed-dependent timbre changes.

Cutting the attack and filtering: changing what a sound "is"

Schaeffer’s cloche coupée (cut bell): remove a bell’s attack and it sounds like an oboe or an organ. The attack carries most of our source identification; take it away and the sustain becomes an abstract tone. Filtering then reshapes the object’s spectrum.

1. In Simpler (Classic), move the Start marker past the transient, then set a fade-in (Simpler’s Fade In / amp envelope Attack 40–200 ms). The bell now begins in its sustain.
2. In an audio clip, the same with clip fades on, or with Warp on in Beats mode: Beats mode’s Transient Envelope control (with Preserve set to transients) shortens each segment’s decay — set it low to chop sounds into dry fragments, or use it on long tones to damp re-attacks.
3. Filtering: EQ Eight with 2–3 narrow bells boosted +12 dB and a low/high cut around them isolates one partial region — a cymbal becomes a whistling tone cluster. Auto Filter (12.2) band-pass with very slow LFO for a moving spectral window.
4. Combine: cut attack + −12 st speed change + band-pass = an object almost impossible to identify. That’s the acousmatic goal.

Tape music: unequal loop lengths (Eno-style systems)

Several loops of different, unrelated lengths, each containing one sound or phrase, played together. Because their lengths share no common factor, the combination changes for a very long time. The tape-loop system behind the classic airport-ambient pieces, rebuilt in Session view.

1. Set the tempo to 60 BPM so lengths read in beats = seconds. Make 5–7 audio clips, each one short sound followed by silence.
2. Set each loop length to a different "prime-ish" value — 17.3, 21.1, 25.9, 31.4, 33.7 s. With Warp off you can set any length; with Warp on, use odd beat values (17.25, 21, 26.5…).
3. Launch all clips at once from one scene. Record the Session output into Arrangement for 30–60 minutes; pick the best passages.
4. Optional: shared reverb and a long Echo send so sounds overlap into a continuous bed.

Reich-style phasing: two clips at slightly different lengths

Two identical loops started together, one very slightly longer (audio) or one step longer (MIDI). They slide out of sync, creating new composite rhythms and melodies, until they eventually come back into unison. The technique of the classic 60s tape-phase pieces, which Live does natively.

1. Stepwise phasing (MIDI): write a 12-note 16th cell. Duplicate the clip to a second track. Clip A loop = 12 sixteenths (0.3.0 at 4/4 = 3 beats). Clip B: same notes, loop = 13 sixteenths. B shifts one step later every cycle; after 12 cycles of B they realign.
2. Continuous phasing (audio): two copies of a speech or instrument loop, Warp off. Make clip B’s loop end a few milliseconds longer (zoom in, nudge the loop end). Or Warp on with clip B’s "Seg. BPM" set 0.2–0.5 BPM different. The drift is gradual — first a flange, then an echo, then a canon.
3. Pan the two voices apart (40–60% L/R) so you hear the composite and the individual lines.
4. Operator marimba: Osc A sine, Osc B ratio 4 (or 3.98 for slight inharmonicity) at −18 dB with a fast-decay envelope; amp decay 300 ms, sustain 0.

Granular synthesis with Granulator III

Granulation cuts a sound into thousands of tiny overlapping windows (grains, 10–200 ms) and reassembles them — freezing time, scrubbing slowly through a recording, or scattering it into a cloud. Granulator III (Max for Live, Suite) can granulate a dropped sample or capture live input.

1. Drop a 10–60 s recording into Granulator III. Start in Classic mode, position at an interesting moment, scan speed at zero: time is frozen, pitch preserved — the granular "time-stop".
2. Grain size: 20–50 ms = gritty, rough; 80–200 ms = smooth, the source’s timbre audible. Spray (position randomisation) 20–60% blurs exactly when grains come from — less spray = more source rhythm.
3. Switch to Cloud mode for dense, asynchronous texture. Play chords on Push: each note is a transposed cloud.
4. Scan very slowly (much slower than real time) to stretch a 30 s recording to 10 min — the granular route to extreme time-stretching.
5. Use the live capture/input option to granulate a live instrument or the modular in real time; MPE (Push 3) maps pressure to grain parameters for expressive control.
6. Modulate position with a slow Max for Live LFO or a Shape-generated MIDI CC so the scan wanders like a reading head.

Spectral composition with Spectral Resonator and Spectral Time

Spectral music treats the harmonic content of a sound as the source of harmony. In Live, the Spectral devices let you impose pitch onto noise, freeze a spectrum into a chord, and shift or smear partials.

1. Pitched noise: Spectral Resonator on a noise-like recording (wind, rain, room tone). Set its MIDI sidechain to a MIDI track playing a sustained chord: the noise rings at those pitches. Decay 2–5 s for drone, short for plucked textures.
2. Use Harmonics/stretch controls to make the resonances inharmonic (bell-like) or compressed; Shift to move the whole partial set; Unison to thicken.
3. Frozen spectra: Spectral Time with Freeze in manual mode — freeze an instant of a sound into a static chord of its partials. Crossfade between successive freezes (set a long fade) for slow harmonic motion derived from the source.
4. Spectral delay: a second Spectral Time in delay mode with Tilt (frequency-dependent delay) and Spray — highs and lows arrive at different times, smearing gestures into spectral waves. Shift in the feedback path makes each repeat climb or fall.
5. Analyse a sound’s main partials (Spectrum device or by ear), write those pitches as a chord into a MIDI clip, and feed that to Spectral Resonator on a different sound — harmony derived from timbre, the core spectralist move.

Microtonality: Live 12 tuning systems and just-intonation drones

12-tone equal temperament makes every interval except the octave slightly out of tune with the harmonic series. Just intonation uses whole-number ratios, so sustained intervals lock without beating — the basis of the long-tone drone tradition. Live 12 makes this practical with per-set tuning systems.

1. In the browser’s Tunings section, choose a just-intonation tuning (or load your own Ableton tuning file, a Scala-based format) and drag it into the set. The tuning applies to Live’s instruments and the clip editor shows the scale degrees; you can set the reference pitch so 1/1 is your drone root.
2. Write a drone chord with long notes. Ratios to listen for: 3/2 (fifth, +2 ct vs ET), 5/4 (major third, −14 ct), 7/4 (harmonic seventh, −31 ct), 9/8 (major second, +4 ct), 11/8 (−49 ct, between fourth and tritone).
3. Use simple waveforms (sine, triangle, filtered saw) and no vibrato/chorus — you want to hear the absence of beating and the combination tones.
4. Without a tuning file: put each drone voice on its own track and detune with the instrument’s fine control by the cent values above (Operator Fine, Analog/Drift detune).
5. Hardware: tune the Buchla 258t clone’s two oscillators by ear to a pure 3:2 or 7:4 — listen for the beating to slow to a stop. Record long takes. This is how drone musicians tuned before tuning tables.

More in tuning, scales & MPE.

Deep listening and long-form drone (Young / Radigue as style)

Drone pieces work with change too slow to notice moment to moment: a partial swelling over three minutes, a filter moving a few hertz per minute, beating patterns shifting as one oscillator drifts by a fraction of a cent. The discipline is to automate slower than feels reasonable.

1. Base tone: Analog with two oscillators in a just interval, slight drift (or Drift with Drift amount 5–15%) for organic slow beating. Envelope: attack 10–30 s, sustain full, release 30 s.
2. Wavetable using a Harmonics-category table: LFO at 0.002–0.01 Hz (one cycle per 2–8 minutes) on wavetable position, so partials fade in and out one by one — the Radigue-style "slow partial" sensation.
3. Draw automation in Arrangement: one filter sweep across 10 minutes, a 3 dB gain swell across 4 minutes. Zoom out and check the curves look almost flat — that’s right.
4. Feedback drone: Echo with feedback 70–90% and a slow filter inside the loop, fed by a tiny input. Or Nautilus with long delay, high feedback and slow modulation as a self-sustaining drone generator you sample.
5. Listen on speakers in a room, not headphones: standing waves and head movement are part of the experience.

Acousmatic spatialisation: trajectories, depth and multichannel

In acousmatic music, where a sound comes from and how it moves is a compositional parameter as important as pitch. Stereo can carry a lot: left/right trajectories, near/far depth, width as "size".

1. Lateral motion: automate Utility’s pan/balance or use Auto Pan-Tremolo (12.3) in pan mode with a very slow, free-running rate. Curved, accelerating trajectories feel physical; linear sweeps feel synthetic.
2. Depth: distance = lower level + less high frequency (EQ Eight high shelf down) + more reverb with less pre-delay ratio. Automate all three together for a sound approaching from far away.
3. Size: width from 0% (a point) to 150% (enveloping) with Utility; reverb early reflections widen too.
4. Multichannel: Live can send tracks to any interface output pairs. For quad or 8-channel, either route groups to output pairs and pan within them, or use a Max for Live surround panner device if your Suite/pack includes one — check the Max for Live section of your browser. Otherwise, build it: one track per speaker pair, a Max for Live LFO or macros crossfading the sends between them.
5. Compose in stereo for streaming; design a separate diffusion version for concerts (see the fixed-media technique).

Field recording composition: listening, editing, layering

From documentary soundscape to fully transformed material, field recording work starts with careful listening and ends with very gentle editing. The principle: respect the recording’s space, then decide how far to move away from it.

1. Record long (10–30 min) at each location; note time, place and gain. 48 kHz/24-bit; higher sample rates (96 kHz) if you plan extreme slow-downs (more ultrasonic content to bring down into hearing).
2. Clean: high-pass 30–60 Hz for wind/handling, notch hum (50/60 Hz and harmonics) with narrow EQ Eight bands, fade every edit (50–500 ms crossfades — hard edits in ambience are very audible).
3. Structure: layer 2–4 recordings with long crossfades; let one dominate each section. Build a "composed soundscape" by placing events (a bird, a door) from other takes at musical moments.
4. Transformation spectrum: documentary (no processing) → enhanced (EQ, light reverb) → tuned (Spectral Resonator keyed to a drone) → abstracted (granular, extreme stretch). Pieces often travel along that spectrum.
5. Live 12.3’s Stem separation (Suite) can sometimes pull voices out of a crowd recording or isolate musical sounds — results on non-music material are variable, try it.

Generative systems: MIDI generators, Follow Actions and modular CV

A generative piece is a system you design and then let run. Live 12 gives you three layers: generators that write material (Seed, Rhythm, Shape, Stacks, Euclidean), Follow Actions that decide which clip plays next, and CV Tools to let a modular make decisions too.

1. Generate material: in a MIDI clip, use the Seed generator (random pitches within the clip’s scale, density control) and the Shape generator (melodic contour) to write 6 different clips of different lengths. Tweak until each is musically acceptable alone.
2. Follow Actions: on each clip set Follow Action A = Other (or Any) and B = Play Again, with Chance A/B around 60/40, and a Follow Action time per clip that’s different (e.g. 3, 5, 7 bars). The system now walks through the clips randomly. Use Jump to send it to a specific "home" clip occasionally.
3. Scene-level Follow Actions (Live 11+) can move the whole piece between states — e.g. sparse → dense → silent.
4. Modular decisions: CV Tools’ CV In receives a random/S&H voltage from the eurorack and maps it to a macro (filter, density, reverb). CV Triggers can send Live events to the modular to clock or reset it. Two systems influencing each other.
5. Leave it running for an hour, record the Session to Arrangement, and listen back for the passages where the system "composed" something. Then refine rules, not notes.

Algorithmic and aleatoric: Random, note Chance and probability

Chance music (in the Cage tradition) uses indeterminacy so the composer decides the space of possibilities, not the result. Live has probability at the note level and in many devices.

1. Note Chance: select notes in the clip, open the Note Chance property in the clip editor and set per-note probability. Downbeat anchors 100%, colours 20–50%, rare events 5–10%.
2. Velocity ranges per note (clip editor) give each hit a random velocity within bounds every pass.
3. Random MIDI effect: Chance sets how often the pitch changes, Choices how many alternatives, Scale snaps to the clip scale. Mode Alt cycles instead of randomising — determinism inside chance.
4. Audio-level chance: Beat Repeat’s Chance; Follow Actions with Chance values; Max for Live LFO in random (S&H) shape on any parameter.
5. Constrain the space: chance music works when the possibilities are all good. Narrow scales, consistent timbre, controlled density.

Prepared-instrument simulation with Collision and Tension

Prepared piano (screws, bolts, rubber between strings) turns pitched strings into small, unpredictable percussion orchestras. Live’s physical models can simulate that: strings with inharmonic stiffness, bouncing hammers, beams and plates excited by mallets.

1. Tension: Exciter Hammer (bouncing) for the rattly double strike of a bolt-prepared string. Increase the damper’s mass/stiffness so notes buzz and choke. Inharmonicity / string stiffness up for a metallic out-of-tune quality. Body type Piano at small size.
2. Collision: Mallet exciter with high stiffness and some noise; Resonator 1 Beam or Plate, Resonator 2 Membrane; Inharmonics +30–60%; Decay short to medium. Different notes on the same patch yield wildly different timbres — exactly like preparations that change by key.
3. "Preparation map": put a Drum Rack of 8 different Collision/Tension patches, each assigned to a key range, so every key is a different preparation. Use one note per preparation like Cage’s tables.
4. Real piano sample route: Simpler with a piano note → Corpus (Beam, short decay) mixed at 20–40% adds a rattle; Erosion Sine at a high frequency adds a buzzing "paper between strings" quality.

More on these models in strings, brass & physical models.

Extreme time-stretching: PaulStretch-like results with Texture/Complex Pro and Spectral Time

Stretching a recording to 20–100 times its length dissolves rhythm and articulation into a smooth, shimmering drone that keeps the source’s harmonic colour. Dedicated tools do this in one pass; in Live you combine warping, resampling and spectral smearing.

1. Warp a 10–30 s recording in Texture mode (grain size 60–120, Flux 30–60% for smooth, randomised grain timing) or Complex Pro (Formants 100%, Envelope ~128 for clean tonal material).
2. In Arrangement, drag the end warp marker so the clip runs at 1–5% speed (a 10 s clip becomes 200–1000 s). Alternatively set the clip’s Seg. BPM very high relative to a 20–60 BPM set tempo, and press the clip’s ÷2 button repeatedly (each press doubles the playback length). Combining both is fastest.
3. Warp artefacts at extreme ratios get "phasey" — resample in stages: stretch 5×, bounce (Bounce in Place, 12.2+), stretch the bounce 5× again. Each pass smooths.
4. Smooth the result with Spectral Time’s delay (high spray, moderate feedback) and long reverb; that recreates the window-overlap smearing of PaulStretch-type algorithms.
5. Granulator III with a near-zero scan speed is an alternative route for live stretching.

Sound-art installations: a Live Set that runs for hours

An installation set must start itself, run for 8–12 hours a day without crashing or drifting into silence or overload, and survive a power cycle.

1. No dead ends: every clip’s Follow Actions must lead to another clip (avoid Stop unless a silence clip with a Follow Action back exists). Test by fast-forwarding: temporarily set Follow Action times to 1 bar and let it run overnight.
2. CPU headroom: Freeze or Flatten anything static (Bounce in Place 12.2, Bounce Group 12.3), raise the buffer to 1024–2048 samples (latency doesn’t matter), aim for a CPU meter under 40%. Disable unused tracks and plugins.
3. Memory: long audio clips set to RAM mode only if the machine has enough memory; otherwise stream from SSD. Avoid devices that record or buffer indefinitely.
4. Safety: Limiter on the Main, ceiling −3 dB; a high-pass on the master for speaker protection; time-of-day volume automation via a long clip envelope if needed.
5. Autostart: at OS level — auto-login, launch Live with the set at startup, disable sleep, screen saver, updates and notifications, and Wi-Fi if not needed. In Live, a startup scene launches on load via a MIDI or Follow Action trick (or a simple Max for Live device that fires a scene on load).
6. Daily cycle: scheduled shutdown/restart at closing time keeps long-term memory and drift issues away.

Performing fixed media: stems, diffusion and Push

Acousmatic pieces are often "fixed media", but in concert they are diffused: a performer at the desk shapes the dynamics and space live, sending stems around a speaker orchestra. Live is an excellent diffusion instrument.

1. Bounce the piece to 4–8 stems by function (low bed, mid bed, foreground events, high details) with Bounce Group (12.3) or by soloing groups, and import them to start-aligned audio tracks in Arrangement.
2. Route each stem to a group per speaker pair, or keep one main output and use sends to extra output pairs for "far", "wide" and "rear" speakers.
3. Map Utility gain of each stem and each send level to Push encoders (or a fader controller). Diffusion is fader work: accentuate gestures, move events around the room, exaggerate dynamics.
4. Stereo-only venue: diffuse width (Utility width per stem), depth (send to a big reverb) and dynamics — still a performance.
5. Rehearse in the venue, at the actual level; set a Limiter per output pair. For a hybrid set, keep one live layer (granular capture, modular through External Audio Effect) on top of the fixed stems.