Speech synthesis and text-to-speech

Machines that produce spoken voice from codes or text; the basis of screen readers, talking devices, voice assistants and today's natural AI voices.

AI & roboticsHardware
  1. 1939

    First version

    Homer Dudley (Bell Labs, Voder); later John Kelly and Carol Lochbaum (Bell Labs), Texas Instruments, Dennis Klatt (MIT, DECtalk) and DeepMind (WaveNet)

    Image: Photographer not credited; image copyright credited to AT&T / Bell Laboratories · Copyright AT&T / Bell Laboratories; sou… · Synthetic Sound Labs visit to the AT&T Archives…
  2. 1961
    Computer-synthesised speech ('Daisy Bell')

    Computer-synthesised speech ('Daisy Bell')

    Kelly and Lochbaum's vocal-tract model on an IBM computer at Bell Labs produces speech and song in software (dated 1961–62).

    Image: ArnoldReinhold · CC BY-SA 3.0 · Wikimedia Commons
  3. 1978
    Speak & Spell (single-chip speech synthesis)

    Speak & Spell (single-chip speech synthesis)

    TI's linear-predictive-coding chip puts digital speech into a cheap consumer toy.

    Image: FozzTexx · CC BY-SA 4.0 · Wikimedia Commons
  4. 1984
    DECtalk text-to-speech

    DECtalk text-to-speech

    Rule-based synthesis reads any typed English text aloud in intelligible voices; built on Klatt's work, from which Stephen Hawking's voice also derived.

    Image: DigitalRampage / Australian Computer Museum Society · Unclear; the Australian Computer Museum… · Australian Computer Museum Society Forum
  5. 2016
    WaveNet neural speech

    WaveNet neural speech

    A deep neural network generates the raw audio waveform itself, closing much of the gap to human-sounding speech.

    Image: MetricSupporter89 · CC BY-SA 4.0 · Wikimedia Commons

See it on the timeline