Glossary
Text to speech (TTS)
Text to speech (TTS), also called speech synthesis, is software that converts written text into spoken audio. Modern systems use neural networks, and they can run on the device or in the cloud.
Updated October 17, 20264 cited facts, each traced to its source
Also called TTS, speech synthesis, voice synthesis.
Text to speech (TTS) is software that converts written text into spoken audio. You give it a sentence, and it returns a synthetic voice reading that sentence aloud. The same field is called speech synthesis or voice synthesis. TTS is the reverse of automatic speech recognition, which turns speech into text. Microsoft's Speech API documentation puts both directions in one interface: it reduces the code an application needs to use both speech recognition and text-to-speech 1.
A TTS system turns spelling into sound in stages
A TTS system first normalizes the text. It expands numbers, abbreviations and symbols into words, so "3" becomes "three" and "Dr." may become "doctor" or "drive" depending on context. Next it converts spelling into pronunciation, which is harder than it looks: the same letters can be said differently, as with "read" in the present and past tense. A neural acoustic model then predicts the pattern of the sound over time, including pitch and pauses. A vocoder turns that pattern into an audio waveform.
Older systems stitched together recorded fragments of a human voice. Newer neural systems generate the audio directly, which is why their voices sound more natural.
A concrete example: a message read aloud
Paste "Dr. Lee moved the 3 pm call to Room 4." into a TTS reader. A good engine expands "Dr." to "doctor" before a name and says the time in words. A weaker engine may guess wrong on the abbreviation, so check that first when you judge a voice.
Why TTS matters for dictation and typing
TTS is the playback half of a dictation workflow. Hearing your own draft is a way to catch errors that your eyes skip over. Dragon's "Read That" feature does this. Nuance says earlier Dragon versions called the feature Text-to-speech 2.
Local TTS can also be fast. Piper, an open neural TTS system, runs well on small hardware. Home Assistant reports that on a Raspberry Pi, medium quality models can generate 1.6 seconds of voice in one second 3. That figure comes from one project on one board, and other models and hardware will differ.
The evidence that hearing text helps reading is strongest for people who struggle to read. A 2018 meta-analysis of students with reading difficulties found an average weighted effect size of 0.35 on reading comprehension for text-to-speech and related read-aloud tools 4. That is a modest gain, and those studies measured comprehension, not proofreading your own draft. Our guide to dictation for blind and visually impaired people covers how spoken output fits into a dictation routine.
Related terms
- Automatic speech recognition (ASR): the opposite direction, speech to text.
- Voice cloning: a TTS voice built to sound like a specific person.
- Voice agent: software that listens and speaks back to finish a task.
- Speech-to-speech model: a model that takes speech in and produces a spoken reply.
Guides that cover it
Evidence
Every number on this page, with where it comes from. Open a fact to see the exact words in its source.
- 1.
- 2.
- 3.
- 4.In a 2018 meta-analysis, text-to-speech and related read-aloud tools had an average weighted effect size of 0.35 on reading comprehension for students with reading difficulties, with a 95 percent confidence interval of 0.14 to 0.56.
Sarah G. Wood et al., Journal of Learning Disabilities, 2018 · Original
Related terms
- Speech to Text AI: Dictation, Assistants and Voice AgentsThe topic this term belongs to
- AI text detectorAn AI text detector is software that estimates whether a passage was written by a language model, usually from the words alone. Detectors make mistakes in both directions, so a score is weak evidence about how a text was made.
- Generative AIGenerative AI is the class of AI models that create new content, such as text, audio or images, from patterns in their training data. The NIST profile defines it as models that emulate the structure and characteristics of input data to generate derived synthetic content.
- Generative error correction (GER)Generative error correction (GER) passes a speech recognizer's output, often its N-best list of alternative transcripts, to a large language model, which rewrites the transcript to fix errors. It can recover words the recognizer never offered, but it can also change words that were already right.
- Intent recognitionIntent recognition is the step that decides what a spoken request is asking for, such as setting a timer or sending a message, and which details the action needs. Voice assistants and voice agents rely on it before they act.
- Large language model (LLM)A large language model (LLM) is a language model with a very large number of parameters, trained on huge amounts of text to predict and generate language. Google's glossary defines the minimum as a language model having a very high number of parameters.
- Machine learningMachine learning is a method in which a program learns its behavior from example data rather than from rules written by hand. Google's glossary defines it as a program or system that trains a model from input data.
- Prompt (AI)In AI, a prompt is the input you give a model to get a response, such as a question, an instruction or pasted text. Google's glossary defines it as any text entered as input to a large language model to condition the model to behave in a certain way.
- Speech language modelA speech language model (speech LLM) is a large language model that accepts audio as input, so it can process speech directly instead of working only on a text transcript. The 2023 SALMONN model is one example that lets an LLM process general audio inputs.
- Speech-to-speech modelA speech-to-speech model takes audio in and produces audio out with one model, without writing the conversation out as text in between. Most voice systems instead chain speech recognition, a language model and speech synthesis.
- Turn detectionTurn detection is the step that decides when a speaker has finished a turn, so a voice system knows when to respond. A pause can be a hesitation rather than an ending, so the decision trades speed against cutting the speaker off.
- Voice agentA voice agent is a voice system that listens, reasons with a language model, uses tools, and speaks back across several turns to finish a task, such as checking an order and rescheduling it. Benchmarks show that agents still fail many tasks on realistic audio.
- Voice assistantA voice assistant is software that listens for a spoken request, works out what you want, and then acts or answers, such as setting a timer or turning off a light. Dictation is different: it turns your speech into text for you to edit.
Pages that use this term
- Speech to Text AI: Dictation, Assistants and Voice AgentsTopic
- Home Assistant Voice: Speech to Text and Voice CommandsGuide
- AI Voice Agents: How They Work and Where They FailGuide
- Can Whisper AI do text to speech?Question
- Can you disable or uninstall Speech Recognition and Synthesis from Google?Question
- What is the best speech to text for Home Assistant?Question
- What is the difference between cascaded and speech-to-speech voice AI?Question
- Speech Services by GoogleGlossary
- Speech-to-speech modelGlossary