Stephen Hawking’s voice, measured

The read

Hawking was offered newer, smoother synthesizers for years and always refused: the robotic DECtalk had become his voice, and he kept it. Measured like any other reading — and flagged as the machine it is — the famous monotone turns out not to be one: melodic range and variability both optimal, pause length optimal, the pause ratio a whisker under its band. None of that is the chip deciding. Hawking composed sentences in advance, word by word, and the synthesizer played the timing he authored — every stop deliberate in the most literal sense possible. The one miss is warmth, reading low: the synthesis, not the man. Pacing was the part of speech he still controlled, and it shows.

Pitch contour · every voiced frame

Pitch & melody

Average pitch133Hz

in the zoneoptimal 100–150 Hz

Pitch variability4.34st

in the zoneoptimal 3–6.5 st

Pitch range8.66st

in the zoneoptimal 7–12 st

Pitch dynamism (PDQ)0.31

in the zoneoptimal 0.15–0.38

Pacing & pauses

Pause ratio16.6%

below the zoneoptimal 18–35 %

Average pause length0.57s

in the zoneoptimal 0.4–1 s

Syllable rate3.76syl/s

below the zoneoptimal 4–5.5 syl/s

Dynamics

Volume dynamics28.5dB

above the zoneoptimal 8–18 dB

Loudness variability8.87dB

above the zoneoptimal 3.5–8 dB

Voice quality

Voice clarity (HNR)10.7dB

below the zoneoptimal 12–30 dB

Pitch steadiness (jitter)2.14%

in the zoneoptimal 0–2.5 %

Volume steadiness (shimmer)12.2%

above the zoneoptimal 0–9 %

Vocal fry (creak)4.0%

in the zoneoptimal 0–8 %

Spectral balance

Presence band (2-4 kHz)0.64%

below the zoneoptimal 6–20 %

Chest band (0-500 Hz)10.2%

below the zoneoptimal 35–75 %

Spectral tilt (alpha ratio)-8.64dB

in the zoneoptimal -22–-8 dB

Energy by band

Low (0–500 Hz)10.2%Mid (0.5–2 kHz)88.8%Presence (2–4 kHz)0.6%Brilliance (4 kHz+)0.4%

The presence band (2–4 kHz, red) is what carries a voice to the back of a room.

Comparisons are based on Resonate's acoustic analysis of publicly available speeches. No endorsement or affiliation is implied.

Source

NASA 50th Anniversary Lecture Series, ‘Why We Should Go Into Space’, 21 Apr 2008 — public-domain NASA recording (archive.org). 150s lecture segment analysed — his DECtalk speech synthesizer, not his own vocal tract.

Watch the NASA lecture

Recording caveat

This is a public-domain NASA recording of Hawking's DECtalk speech synthesizer — an intentional, unusual case for this instrument. Its acoustic profile reflects the synthesizer's design, not a human vocal tract, so treat these readings as a measurement of the device, not the man.

Method

analysis-service pipeline (app/analysis.py, app/benchmarks.py, app/qualities.py) run locally over the segments documented in docs/hero-sources.md. Acoustic metrics only — no transcription, so transcript-dependent metrics (wpm, fillers) are absent and the overall score weighs the acoustic set.

Think a measurement is wrong, or want this page removed? Request a correction or takedown — we’ll re-measure it, correct it, or take the page down.

Steal this

the full stop

Even here, what the measurements endorse is the stopping — pauses that exist because they were, literally, written. Compose your next important point as if every pause had to be typed: decide where the stops go before you speak, then honour them at full length. A pause you planned reads as thought. A pause you fell into reads as losing the thread.

From the training library: The full stop

Same instrument · same scale

You’ve just read Hawking’s voice. Now read your own.

Measure my voice

Free · about twenty seconds · no signup. The same six scores on the same scale — and the recording is deleted the moment it’s analysed.