Speech synthesis
Artificial production of human speech by computer systems.
Speech synthesis is the artificial production of human speech, implemented by computer systems called speech synthesizers. These systems convert text or symbolic linguistic representations into spoken output, with applications ranging from accessibility tools to entertainment. A text-to-speech system, or TTS, transforms normal language text into speech, while other systems render phonetic transcriptions or similar linguistic notations into audible form. The reverse process is known as speech recognition. Synthesized speech can be generated by concatenating pre-recorded speech segments stored in a database. The size of these stored units varies; systems using phones or diphones offer the broadest output range but may sacrifice clarity, whereas storing entire words or sentences allows for high-quality output in specific domains. Alternatively, a synthesizer can model the human vocal tract and voice characteristics to produce a completely synthetic voice. The quality of a speech synthesizer is judged by its similarity to natural human speech and its intelligibility. An intelligible TTS program enables people with visual impairments or reading disabilities to hear written text on a computer. The earliest computer operating system to include a speech synthesizer was Unix in 1974, via its speak utility. Later, in 2000, Microsoft Sam became the default TTS voice for the narrator accessibility feature in Windows 2000 and Windows XP. A typical TTS engine comprises two parts: a front-end and a back-end. The front-end first normalizes raw text, converting symbols like numbers and abbreviations into written-out words, a process called text normalization or tokenization. It then assigns phonetic transcriptions to each word and marks prosodic units such as phrases and sentences. This symbolic linguistic representation is passed to the back-end, or synthesizer, which converts it into sound, sometimes computing target prosody like pitch contour and phoneme durations. The history of speech synthesis predates electronics, with early attempts to build mechanical speaking machines, including models of the human vocal tract that produced vowel sounds in the 18th century and later devices that added tongue and lip models for consonants. In the 1930s, Bell Labs developed the vocoder, which analyzed speech into fundamental tones and resonances, leading to the keyboard-operated Vode
- field
- Computer science, linguistics, electrical engineering
- key_technologies
- Text-to-speech (TTS), concatenative synthesis, formant synthesis, linear predictive coding (LPC), line spectral pairs (LSP), deep learning (WaveNet)
- first_computer_TTS_system
- 1968 (Noriko Umeda et al., Japan)
Lore & Background
Speech synthesis involves the artificial generation of human speech, typically by a computer system known as a speech synthesizer, which can exist as either software or hardware. The most common form is a text-to-speech (TTS) system, which converts standard text into spoken words, though other systems may start from symbolic linguistic representations like phonetic transcriptions. The quality of a synthesizer is measured by how closely it resembles a human voice and how clearly it can be understood. Synthesized speech can be produced by concatenating pre-recorded speech units stored in a database; systems that store small units like phones or diphones can produce the widest range of output, though they may sacrifice clarity, while systems storing entire words or sentences for specific domains achieve higher quality. Alternatively, a synthesizer can model the vocal tract and other human voice characteristics to generate a completely synthetic voice. A TTS system has two main parts: a front-end that normalizes raw text (converting numbers and abbreviations into written words) and assigns phonetic transcriptions and prosodic markers, and a back-end synthesizer that converts this symbolic representation into sound, often computing target pitch and phoneme durations. Historically, mechanical attempts to emulate speech date back centuries, with early notable devices including Christian Gottlieb Kratzenstein’s 1779 models of the human vocal tract that produced vowel sounds, and Wolfgang von Kempelen’s 1791 bellows-operated machine that added models for consonants. In the 1930s, Bell Labs developed the vocoder, leading to Homer Dudley’s keyboard-operated Voder exhibited in 1939. The first computer-based speech synthesis appeared in the late 1950s, with Noriko Umeda et al. creating the first general English TTS system in 1968. Linear predictive coding, developed from 1966 onward, became the basis for early synthesizer chips like those in the 1978 Speak & Spell toys.
Reader's Guide
Speech synthesis has evolved from mechanical vocal tract models to sophisticated digital systems. Its significance lies in enabling communication for people with visual impairments or reading disabilities, as text-to-speech programs allow them to listen to written words. The technology also powers virtual assistants, navigation systems, and accessibility features in operating systems. Key milestones include the first computer-based systems in the late 1950s, the development of LPC and LSP coding methods that improved compression and quality, and the shift to deep learning with WaveNet in 2016. The field continues to advance, with modern systems achieving high intelligibility and naturalness. The legacy of speech synthesis is evident in its widespread adoption across consumer electronics, from toys like Speak & Spell to operating system narrators like Microsoft Sam. The technology's history reflects a convergence of linguistics, signal processing, and artificial intelligence, with ongoing research into more expressive and efficient synthesis.
Frequently Asked Questions
What is Speech synthesis?
Speech synthesis is the artificial generation of human-sounding voice by computer systems called speech synthesizers, which transform written text or symbolic linguistic data into audible spoken output.
What key technologies power Speech synthesis?
The field relies on a range of methods including text-to-speech engines, concatenative and formant synthesis, linear predictive coding, line spectral pairs, and modern deep-learning models such as WaveNet.
When was the first computer-based Speech synthesis system created?
The first computer TTS system was demonstrated in 1968 by Noriko Umeda and her colleagues in Japan.
Which academic disciplines does Speech synthesis span?
It sits at the crossroads of computer science, linguistics, and electrical engineering, drawing expertise from all three to turn language into sound.
Why is Speech synthesis important in everyday communication and media?
Its applications range from accessibility tools that help people with visual or speech impairments navigate the world to entertainment and interactive media experiences.
More in Communication And Media 1-21
Spotted an error? Know more?
This is a living reference — every entry is fact-audited, and reader corrections feed straight into our audit queue. Suggest an edit · See this site's audit record