Lucier "I Am Sitting in a Room" Simulation — User Guide
Physical acoustic simulation: recreates the iconic Alvin Lucier piece through iterative convolution with a physically-modeled room impulse response, gradually transforming speech into resonant room tones.
What this does
This script implements physical acoustic simulation of Alvin Lucier's iconic piece — a computational recreation of the 1969 experimental composition "I Am Sitting in a Room" through iterative convolution with a physically-modeled room impulse response. The process: (1) Physical Room Modeling: Generates a synthetic impulse response based on acoustic physics (RT60 decay, reflection density, microphone placement). (2) Iterative Convolution: Repeatedly convolves an audio source (typically speech) with the room model, simulating re-recording in the same space multiple times. (3) Gradual Transformation: With each iteration, speech intelligibility decreases while room resonances become dominant. (4) Physical Parameter Control: Models real acoustic phenomena like microphone proximity effect, reflection decay, and pre-delay. The result is a transformation from clear speech to pure resonant tones, exactly as in Lucier's original conceptual piece.
Key Features:
- Physically-Modeled Room — Synthetic impulse response based on acoustic physics
- Microphone Placement Simulation — Models proximity effect and direct/reflected sound balance
- Iterative Degradation — Gradually transforms speech into room resonances
- Acoustic Parameter Control — RT60 decay time, reflection density, pre-delay
- Realistic Convolution — Uses Praat's convolution for accurate room simulation
- Progress Monitoring — Real-time feedback on iteration progress and peak levels
- Educational Implementation — Makes Lucier's conceptual piece accessible and tunable
- Batch Processing — Automates the tedious manual re-recording process
What is "I Am Sitting in a Room"? Original (1969): Alvin Lucier records himself speaking, plays recording back in room, records that, repeats until speech becomes pure resonance. This simulation: Digital recreation of same process using convolution with synthetic room impulse response. Advantages: (1) Accessibility: Experience Lucier's concept without physical room/tape. (2) Control: Precisely tune acoustic parameters. (3) Repeatability: Consistent results from same parameters. (4) Speed: Minutes instead of hours of manual recording. (5) Educational: Understand acoustic principles through simulation. Use cases: Music education (teaching Lucier's piece), acoustic research (room effect simulation), sound art (creating similar pieces), audio forensics (understanding iterative degradation), composition (exploring room resonance effects).
Technical Implementation: The script creates a physics-based simulation: (1) Impulse Response Generation: Creates synthetic room response with direct sound (pre-delay + proximity gain) and reflections (random timing, exponential decay). (2) Convolution Engine: Uses Praat's convolution to simulate sound propagation through room. (3) Iterative Loop: Repeats convolution 30 times (configurable), cropping to original duration each iteration. (4) Normalization Control: Prevents runaway gain while allowing characteristic Lucier buildup. (5) Physical Modeling: RT60 decay coefficient = 6.9078/RT60 (natural exponential decay), reflection amplitude = (1 - proximity_gain) × exp(-time × decay). Key insight: The simulation mathematically models what happens physically — each iteration represents another generation of tape recording in the same acoustic space.
Quick start
- In Praat, select exactly one Sound object (speech works best).
- Run script… →
lucier_room_simulation.praat. - Set IR_duration_seconds (1.0-2.0s for typical rooms).
- Set RT60_seconds (0.5s = dry room, 2.0s = reverberant).
- Set Number_of_reflections (500-2000, higher = denser reverb).
- Set Microphone_proximity_gain (0.85-0.98, controls degradation speed).
- Set Number_of_iterations (15-40, more = more transformation).
- Click OK — watch console as room model generates and iterations run.
- Final result named "originalname_iteration_X" appears and plays automatically.
Quick tip: For authentic Lucier effect, use speech audio (15-60 seconds). Set Microphone_proximity_gain = 0.90-0.92 for classic degradation rate. Use RT60 = 1.0-1.5 seconds for typical living room. Start with 30 iterations — this matches Lucier's original (~32 generations). Watch the console output — shows impulse response details and iteration progress. The peak level indicator shows how energy builds up (should stabilize around normalization_level). Processing time depends on audio length and iterations — 30s audio × 30 iterations ≈ 1-2 minutes. The output is the final iteration — you can examine intermediate iterations in Objects window before cleanup.
Important: SPEECH WORKS BEST — Lucier's piece uses speech; music/noise may not transform as characteristically. Microphone proximity is critical — Lower values (0.85) cause faster degradation, higher (0.98) cause slower transformation. RT60 affects resonance frequencies — Longer RT60 emphasizes room modes more strongly. Convolution is computationally intensive — Long audio (5+ minutes) with many iterations may be slow. Reflection count affects density — Too few (<200) creates sparse, metallic sound; too many (>5000) may be heavy. Pre-delay simulation models distance from speaker to mic (0.01s ≈ 3.4 meters). Normalization prevents clipping but also affects characteristic Lucier buildup — 0.99 works well. Physical model approximation — Synthetic impulse response, not measured from real room.
The Original Piece
🎵 Alvin Lucier's "I Am Sitting in a Room" (1969)
Concept: Speech gradually transforms into pure room resonance through repeated recording
Original process: Record speech → Play in room → Record result → Repeat ~32 times
Result: Intelligible speech → Rhythmic patterns → Pure resonant tones
Significance: Landmark of conceptual music, acoustic phenomenology, process art
Duration: Original ~45 minutes, this simulation creates shorter versions
Lucier's Original Text
ORIGINAL SCRIPT (spoken by Lucier):
"I am sitting in a room different from the one you are in now.
I am recording the sound of my speaking voice and I am going to
play it back into the room again and again until the resonant
frequencies of the room reinforce themselves so that any semblance
of my speech, with perhaps the exception of rhythm, is destroyed.
What you will hear, then, are the natural resonant frequencies
of the room articulated by speech. I regard this activity not
so much as a demonstration of a physical fact, but more as a
way to smooth out any irregularities my speech might have."
KEY ELEMENTS IN SIMULATION:
1. "play it back into the room again and again" → Iterative convolution
2. "resonant frequencies of the room reinforce themselves" → Room mode emphasis
3. "any semblance of my speech... is destroyed" → Speech intelligibility loss
4. "natural resonant frequencies of the room articulated by speech" → Final result
Original vs Simulation Comparison
| Aspect | Original (1969) | This Simulation | Fidelity |
|---|---|---|---|
| Room | Physical space (various) | Synthetic impulse response | Physical model approximation |
| Recording medium | Analog tape (generation loss) | Digital convolution (perfect) | Different artifact character |
| Process | Manual re-recording (~32×) | Automated iteration (configurable) | Same conceptual process |
| Time required | Hours (real-time) | Minutes (computation) | Much faster |
| Control | Fixed by physical setup | Tunable parameters | More experimental control |
| Result character | Specific room resonances | Parameter-controlled resonances | Conceptually similar |
Educational Context
Teaching Lucier's Concepts:
- Room modes: How spaces emphasize certain frequencies
- Iterative processes: Small changes accumulate dramatically
- Speech vs resonance: Source vs medium characteristics
- Conceptual music: Idea as primary musical material
- Acoustic phenomenology: Physical properties as compositional elements
- Process art: System/process as artwork
Room Physics Model
🔬 Physical Acoustic Simulation
Impulse Response: Mathematical representation of how room responds to sound
Direct Sound: Straight path from speaker to microphone (pre-delay + gain)
Reflections: Sound bouncing off surfaces (random timing + exponential decay)
RT60 Decay: Time for sound to decay 60 dB (reverberation characteristic)
Microphone Proximity: Balance between direct and reflected sound
Impulse Response Generation Algorithm
STEP 1: CREATE EMPTY IMPULSE RESPONSE
Create Sound from formula: "IR", 1, 0, IR_duration, sample_rate, "0"
total_samples = duration × sample_rate
STEP 2: ADD DIRECT SOUND
direct_sample = round(pre_delay × sample_rate)
Set value at sample: direct_sample, microphone_proximity_gain
→ Models sound traveling directly from speaker to mic
STEP 3: CALCULATE DECAY COEFFICIENT
decay_coefficient = 6.9078 / RT60
→ 6.9078 = ln(1000) [60 dB = 1000× amplitude reduction]
→ Natural exponential decay: amplitude = exp(-time × decay_coefficient)
STEP 4: GENERATE REFLECTIONS
reflection_max_amplitude = 1.0 - microphone_proximity_gain
For i = 1 to number_of_reflections:
random_time = pre_delay + randomUniform(0, IR_duration - pre_delay)
time_delta = random_time - pre_delay
natural_decay = exp(-time_delta × decay_coefficient)
random_amplitude = randomGauss(0, 0.3) × natural_decay × reflection_max_amplitude
sample_index = round(random_time × sample_rate)
Add random_amplitude to sample sample_index
STEP 5: NORMALIZE
Scale peak: 0.9 (prevents distortion in convolution)
Physical Parameter Explanations
| Parameter | Physical Meaning | Typical Values | Effect on Result |
|---|---|---|---|
| IR_duration_seconds | Length of room response captured | 1.0-2.0 s | Longer = more reverb tail, more resonance detail |
| RT60_seconds | Reverberation decay time (60 dB) | 0.3-3.0 s | Longer = slower decay, stronger resonance buildup |
| Number_of_reflections | Density of room reflections | 500-5000 | Higher = denser, smoother reverb; lower = sparse, metallic |
| Pre_delay_seconds | Time for direct sound to reach mic | 0.005-0.05 s | Models speaker-mic distance (0.01s ≈ 3.4m) |
| Microphone_proximity_gain | Direct vs reflected sound balance | 0.80-0.98 | Higher = more direct sound, slower degradation |
Microphone Proximity Physics
PROXIMITY EFFECT MODEL:
Direct sound amplitude = microphone_proximity_gain
Reflected sound maximum amplitude = 1.0 - microphone_proximity_gain
PHYSICAL INTERPRETATION:
• Mic very close to speaker (0.97-0.99):
- Direct sound dominates (97-99%)
- Reflections weak (1-3%)
- Slow degradation (many iterations needed)
- Like Lucier with mic close to speaker
• Mic at moderate distance (0.90-0.95):
- Balanced direct/reflected
- Medium degradation rate
- Typical "room recording" sound
• Mic far from speaker (0.80-0.85):
- Reflections dominate (15-20%)
- Direct sound weak
- Fast degradation (speech disappears quickly)
- Like recording from across room
MATHEMATICAL EFFECT IN ITERATIONS:
Each iteration multiplies frequency response by room's response
Frequencies at room resonances get multiplied by >1, grow
Frequencies at room nulls get multiplied by <1, decay
Proximity controls initial balance between original and room-modified signal
RT60 and Room Mode Emphasis
RT60 EFFECT ON ROOM MODES:
Room modes (resonant frequencies) occur at:
f_n = (n × c) / (2 × L) for dimension L, speed of sound c
With RT60 decay: amplitude(t) = exp(-t × decay_coefficient)
decay_coefficient = 6.9078 / RT60
IN ITERATIVE PROCESS:
Frequency response at iteration i: H(f)^i
Where H(f) is room's frequency response
Room modes (resonances) have |H(f)| > 1
With each iteration: resonance_gain = |H(f_resonance)|^i
RT60 affects how peaked resonances are:
• Short RT60 (0.5s): Broad resonances, many frequencies emphasized
• Medium RT60 (1.0s): Moderate resonance peaks
• Long RT60 (2.0+ s): Sharp, narrow resonances
EXAMPLE CALCULATION:
RT60 = 1.0 s → decay_coefficient = 6.9078
Resonance at 100 Hz might have H(100) = 1.2
After 30 iterations: gain = 1.2^30 ≈ 237
That frequency becomes 237× louder relative to others
Parameter Guide
Room Acoustic Parameters
| Parameter | Default | Minimum | Maximum | Acoustic Equivalent | Effect on Transformation |
|---|---|---|---|---|---|
| IR_duration_seconds | 1.5 | 0.5 | 5.0 | Reverb tail length | Longer = more complex resonance patterns |
| RT60_seconds | 1.0 | 0.3 | 5.0 | Reverberation time | Longer = stronger, narrower resonances |
| Number_of_reflections | 1000 | 100 | 10000 | Reflection density | Higher = smoother, more diffuse sound |
| Pre_delay_seconds | 0.01 | 0.001 | 0.10 | Speaker-mic distance | Longer = more distinct early reflections |
Recording Setup Parameters
| Parameter | Default | Lucier-like | Fast Degradation | Slow Evolution | Physical Meaning |
|---|---|---|---|---|---|
| Microphone_proximity_gain | 0.95 | 0.90-0.92 | 0.80-0.85 | 0.97-0.99 | Direct/refl. balance |
| Number_of_iterations | 30 | 25-35 | 15-20 | 40-50 | Generations of recording |
| Normalization_level | 0.99 | 0.98-1.00 | 0.95-0.98 | 0.99-1.00 | Peak level control |
Preset Configurations
🏠 Classic Lucier Configuration
Goal: Recreate characteristic sound of original piece
Parameters:
- IR_duration_seconds: 1.2-1.5
- RT60_seconds: 1.0-1.2
- Number_of_reflections: 800-1200
- Pre_delay_seconds: 0.01
- Microphone_proximity_gain: 0.90-0.92
- Number_of_iterations: 30
- Normalization_level: 0.99
Result: Speech gradually dissolves into clear room resonances over ~30 iterations
⏱️ Fast Transformation
Goal: Quick demonstration of the concept
Parameters:
- IR_duration_seconds: 2.0
- RT60_seconds: 1.5
- Number_of_reflections: 2000
- Microphone_proximity_gain: 0.85
- Number_of_iterations: 15-20
- Normalization_level: 0.97
Result: Speech disappears quickly, strong resonances emerge by iteration 10
🔍 Detailed Examination
Goal: Study gradual transformation in detail
Parameters:
- IR_duration_seconds: 1.0
- RT60_seconds: 0.8
- Number_of_reflections: 500
- Microphone_proximity_gain: 0.97
- Number_of_iterations: 40-50
- Normalization_level: 0.995
Result: Very slow transformation, allows hearing each stage clearly
🏛️ Cathedral Effect
Goal: Create large, reverberant space sound
Parameters:
- IR_duration_seconds: 3.0-4.0
- RT60_seconds: 2.5-3.5
- Number_of_reflections: 3000-5000
- Pre_delay_seconds: 0.03-0.05
- Microphone_proximity_gain: 0.88-0.92
- Number_of_iterations: 25-30
Result: Rich, complex resonances with long decay tails
Parameter Interaction Guide
KEY INTERACTIONS:
1. PROXIMITY × ITERATIONS:
High proximity (0.97) + Many iterations (40) = Slow, detailed transformation
Low proximity (0.85) + Few iterations (15) = Quick, dramatic transformation
2. RT60 × REFLECTION COUNT:
High RT60 (2.0) + High reflections (3000) = Smooth, singing resonances
Low RT60 (0.5) + Low reflections (300) = Metallic, percussive resonances
3. IR_DURATION × NORMALIZATION:
Long IR (3.0) + Low normalization (0.95) = May clip, reduce normalization
Short IR (1.0) + High normalization (0.99) = Safe, conservative
4. PRE_DELAY × PROXIMITY:
Long pre-delay (0.03) + High proximity (0.96) = Distinct early reflections
Short pre-delay (0.005) + Low proximity (0.87) = Integrated direct/reflected
Simulation Process
🔄 Iterative Convolution Algorithm
Step 1: Generate physical room impulse response
Step 2: Convolve source audio with room response
Step 3: Crop to original duration (simulate fixed recording length)
Step 4: Normalize to prevent clipping
Step 5: Repeat Steps 2-4 for specified iterations
Step 6: Output final iteration as result
Iteration-by-Iteration Transformation
MATHEMATICAL MODEL:
Let:
x(t) = original speech signal
h(t) = room impulse response
* = convolution operator
y_i(t) = signal after i iterations
Process:
y_0(t) = x(t) (original)
y_1(t) = [x * h](t) (first recording)
y_2(t) = [y_1 * h](t) = [x * h * h](t)
...
y_i(t) = [x * h^i](t) (after i iterations)
In frequency domain (convolution → multiplication):
Y_i(f) = X(f) × H(f)^i
Where H(f) is room's frequency response
Room resonances: frequencies where |H(f)| > 1
Room nulls: frequencies where |H(f)| < 1
EFFECT OVER ITERATIONS:
• Resonant frequencies grow: |H(f_res)|^i → large amplification
• Null frequencies decay: |H(f_null)|^i → near silence
• Speech spectrum × resonance pattern → speech shaped by room
• After many iterations: only resonances remain
Transformation Timeline (30 Iterations)
| Iterations | Speech Intelligibility | Room Resonance | Characteristic Sound | Listening Focus |
|---|---|---|---|---|
| 1-5 | 95-80% | Faint | Speech with reverb | Speech content, reverb character |
| 6-10 | 80-50% | Emerging | Echoey, resonant speech | Resonance frequencies appearing |
| 11-15 | 50-20% | Dominant | Speech fragments in resonance | Rhythm vs resonance balance |
| 16-20 | 20-5% | Strong | Pitched tones with speech rhythm | Tonal qualities, rhythmic patterns |
| 21-25 | 5-1% | Very strong | Pure tones, some speech contour | Specific resonance frequencies |
| 26-30 | <1% | Complete | Room resonances only | Acoustic character of synthetic room |
Normalization Strategy
NORMALIZATION PURPOSE:
Prevent digital clipping during iterative gain buildup
Mimic tape saturation/level control in original
ALGORITHM:
After each iteration i:
peak_i = maximum absolute amplitude of y_i(t)
If peak_i > normalization_level:
y_i(t) = y_i(t) × (normalization_level / peak_i)
Else:
No change (preserve natural dynamics)
EFFECT ON LUCIER CHARACTER:
• normalization_level = 1.0: No limiting, may clip (authentic to tape saturation)
• normalization_level = 0.99: Gentle limiting, prevents clipping
• normalization_level = 0.95: Strong limiting, reduces dynamic range
• Lower values flatten dynamics faster
COMPARISON TO ORIGINAL:
Original used analog tape → natural soft clipping
This uses digital normalization → controlled peak limiting
Similar effect: prevents infinite gain buildup at resonances
Cropping to Original Duration
WHY CROP AFTER EACH ITERATION?
Physical analogy: Each recording uses same length tape
Mathematical necessity: Convolution lengthens signal
Duration(y_i) = Duration(x) + Duration(h) - 1
Without cropping:
Iteration 1: duration = D + IR - 1
Iteration 2: duration = (D + IR - 1) + IR - 1 = D + 2×IR - 2
...
Iteration i: duration = D + i×IR - i → grows without bound
With cropping (to original duration D):
Each iteration same duration D
Simulates fixed-length tape recording
Truncates reverb tail beyond D
Focuses on steady-state resonance buildup
PRACTICAL EFFECT:
• Early iterations: lose some reverb tail
• Later iterations: resonance patterns stabilize within D
• Consistent with Lucier's fixed-duration tape loops
Creative Applications
Music Education and Analysis
🏫 Teaching Acoustic Concepts
Goal: Demonstrate room acoustics and iterative processes
Workflow:
- Record student speaking Lucier's text or other material
- Process with different parameter sets
- Compare results: dry vs 5 vs 15 vs 30 iterations
- Analyze frequency spectra at each stage
- Discuss physical vs perceptual transformation
Learning outcomes: Room modes, resonance, iterative systems, conceptual art
Sound Design and Film
🎬 Creating Evolving Soundscapes
Goal: Generate evolving ambient textures for media
Workflow:
- Use non-speech sources (field recordings, instruments)
- Experiment with extreme parameters (RT60=4.0, proximity=0.99)
- Process for many iterations (40-50)
- Use resulting tones as atmospheric beds
- Layer with original or other processed versions
Applications: Sci-fi atmospheres, dream sequences, transitional sounds
Composition and Sound Art
🎵 Lucier-inspired Compositions
Goal: Create new works using Lucier's process with variations
Workflow:
- Use different source material (poetry, singing, instruments)
- Experiment with non-room impulse responses
- Combine with other processing (filtering, time-stretching)
- Create multi-movement piece with different parameter sets
- Perform live with real-time parameter adjustment
Creative variations: Different "rooms," multiple sources, hybrid processes
Acoustic Research
🔬 Room Mode Visualization
Goal: Study room resonance patterns through iterative emphasis
Workflow:
- Use broadband noise or impulse as source
- Process with high iteration count (50+)
- Analyze spectrum of result
- Compare to mathematical room mode predictions
- Visualize resonance buildup over iterations
Research applications: Room acoustic analysis, resonance identification, modal decay studies
Advanced Creative Techniques
Multi-Room Processing:
- Process source with Room A parameters (e.g., small room)
- Take result, process with Room B parameters (e.g., large hall)
- Creates hybrid acoustic space
- Example: Speech → Small room (10 iterations) → Cathedral (20 iterations)
Iteration Extraction:
- Modify script to save all iterations (not just final)
- Create piece using different iterations as sections
- Example: Movement 1 (iteration 5), Movement 2 (iteration 15), etc.
- Shows transformation process as musical structure
Hybrid Sources:
- Process multiple sources separately
- Mix results at different stages
- Example: Speech transforms to resonance, then mix with transforming music
- Creates complex interplay of dissolving materials
Troubleshooting Common Issues
Problem: Speech doesn't transform (stays intelligible)
Causes: Microphone_proximity_gain too high (>0.97), iterations too few (<15)
Solutions: Decrease proximity (0.85-0.92), increase iterations (25-35), increase RT60 (1.2-1.5)
Causes: Microphone_proximity_gain too high (>0.97), iterations too few (<15)
Solutions: Decrease proximity (0.85-0.92), increase iterations (25-35), increase RT60 (1.2-1.5)
Problem: Result clips/distorts
Causes: Normalization_level too high (1.0), strong resonances overwhelming
Solutions: Decrease normalization (0.95-0.98), decrease RT60 (0.8-1.0), decrease proximity (more reflections)
Causes: Normalization_level too high (1.0), strong resonances overwhelming
Solutions: Decrease normalization (0.95-0.98), decrease RT60 (0.8-1.0), decrease proximity (more reflections)
Problem: Result is too noisy/metallic
Causes: Number_of_reflections too low (<300), RT60 too short (<0.5)
Solutions: Increase reflections (1000-2000), increase RT60 (1.0-1.5), increase IR_duration (1.5-2.0)
Causes: Number_of_reflections too low (<300), RT60 too short (<0.5)
Solutions: Increase reflections (1000-2000), increase RT60 (1.0-1.5), increase IR_duration (1.5-2.0)
Problem: Processing takes too long
Causes: Long source audio, many iterations, high reflection count
Solutions: Use shorter source (30-60s), decrease iterations (20-25), decrease reflections (500-800)
Causes: Long source audio, many iterations, high reflection count
Solutions: Use shorter source (30-60s), decrease iterations (20-25), decrease reflections (500-800)
Performance Guidelines
| Scenario | Source Length | Iterations | Reflections | Processing Time |
|---|---|---|---|---|
| Quick demo | 10-20s | 15-20 | 500-800 | 30-60 seconds |
| Standard Lucier | 30-60s | 25-30 | 800-1200 | 1-3 minutes |
| Detailed study | 60-120s | 35-40 | 1500-2000 | 3-8 minutes |
| Extreme processing | 120s+ | 40-50 | 3000-5000 | 10+ minutes |