What is Text-to-Speech (TTS)?

Text-to-speech, also known as TTS, is a technology that converts written words into audible speech. An AI voice generator communicates with users when reading a screen is impossible or inconvenient. Text-to-speech technology opens up applications and information to be used in new ways, improving accessibility for individuals who cannot read text on a screen. \n

Text-to-speech technology has evolved over the last few decades. Deep learning makes it possible to produce very natural-sounding speech that includes pitch, rate, pronunciation, and inflection changes. Today, computer-generated speech is used in various use cases and is becoming ubiquitous in user interfaces. Newsreaders, gaming, public announcement systems, e-learning, telephony, IoT apps and devices, and personal assistants are just starting points.","id":"collection-text-media#text-to-speech-1","customSortOrder":"1","heading":"What is text-to-speech?"},"metadata":{"tags":[]}},{"fields":{"bodyContent":"

Speech synthesis makes applications more accessible, allowing users to consume and comprehend information without having to focus on a screen. Here is a quick overview of some key advantages to using text-to-speech technology. \n

Accessibility \n

Text-to-speech caters to various communication styles and preferences, making digital content accessible to a broader audience. It improves access for users who cannot read due to visual impairments, literacy challenges, age, or other health concerns. As an assistive technology, it offers an alternative way to get information and ensure inclusivity. \n

Enhanced learning \n

Text-to-speech is applied to online materials to facilitate e-learning. Combining visual and audio presentations improves comprehension, recall, vocabulary skills, motivation, and confidence. The technology reads digital text aloud so language learners can understand how to pronounce words and phrases accurately. Hearing the text also reinforces vocabulary retention and sentence structure understanding. \n

Mobility & freedom \n

Text-to-speech can turn any digital content into a multimedia experience. People can listen to news, blog articles, or even a PDF document on the go or while multitasking. Flexibility boosts productivity as users consume content hands-free. \n

Engagement and user experience \n

TTS technology encourages users to engage with lengthy articles, reports, or books. They can access more written content in less time, improving content retention. It improves application metrics like visitor count and time spent on site. You get more conversions by enhancing the customer journey. \n

Fast and affordable \n

Cloud computing has made it fast and easy to implement text-to-speech. The cloud's economics of scale also make it inexpensive to integrate. You don't have to pay upfront or minimum monthly fees to start. You only pay if and when users access the feature.","id":"collection-text-media#text-to-speech-2","customSortOrder":"2","heading":"What are the benefits of text-to-speech?"},"metadata":{"tags":[]}},{"fields":{"bodyContent":"

Applications that use voice to communicate are becoming more common every day. With text-to-speech solutions, your websites, mobile apps, digital books, e-learning tools, and online documents can literally have their own voice. We give some example use cases below. \n

Audio publishing \n

Publishers and content owners can quickly and inexpensively convert books, articles, and written material into audio with text-to-speech. You can convert existing written text to target a broad learner base for e-learning and training use cases. Turn your content into a more effective and less costly format to roll out across multiple languages. \n

Customer service \n

TTS systems enhance the quality of interactive call centers and support communication applications. Build better chatbots and AI assistants that read aloud digital text for users when requested. It is also a key technology in interactive response mechanisms and automated phone systems. Extend automated customer service interactions beyond monotonous phrases to conversational responses that feel empathetic and improve customer satisfaction. \n

Media & entertainment \n

TTS technology can be used to generate voiceovers for videos, animations, and interactive games. It lowers costs and increases efficiency in media pre-production and development. It also allows for real-time narration and dynamic commentary based on player actions in gaming or interactive apps. You can also use text-to-speech tools to deliver immersive audio content in virtual reality (VR) environments. \n

Healthcare \n

TTS technology in healthcare opens communication lines with patients and addresses the shortage of healthcare professionals. Generative AI-powered applications with voice interfaces can interpret patient queries and intent, triage patients, and respond in a natural-sounding voice. They can do everything from booking appointments to supporting treatment management and medicine reminders without forcing the patient to read a screen.","id":"collection-text-media#text-to-speech-3","customSortOrder":"3","heading":"What are the use cases of text-to-speech technology?"},"metadata":{"tags":[]}},{"fields":{"bodyContent":"

Text-to-speech systems use powerful artificial intelligence (AI) and machine learning (ML) models to generate spoken words from text. The models run on deep neural networks—computing nodes that link and work together, like the human brain. The deep neural networks are trained on voice data in various languages, accents, pitch, and volume. During training, both the audio clip and its corresponding transcribed text are given to the AI model. The model identifies co-relations and patterns between the written and spoken text. It uses that knowledge to analyze and convert new text to sound. \n

The process works as follows. \n

Transforming text into time-aligned features \n

The neural network first takes the input text and converts it into time-aligned features that represent the detailed characteristics of speech over time, such as pitch, rhythm, and tone. Common features include: \n

\n
• Mel spectrogram, which shows how sound frequencies change over time. \n
• F0 frequencies, which represent the pitch or fundamental speech frequency. \n \n

  The system also considers linguistic features like how certain sounds should be pronounced or stressed, aligning them with the timing needed to make the speech sound natural. For example, the word hello has a short first sound followed by a long second sound. \n

  Converting time-aligned features into audio \n

  The next step transforms these features into audio that sounds like human speech. The neural network processes the features to synthesize a smooth and natural-sounding voice. Advanced text-to-speech technologies give you features like: \n

  \n
  • Volume control with whisper capability \n
  • High and low-pitch \n
  • Fast or slow speed. \n
  • Multiple languages and accents \n
  • Multiple speaking styles, including customized voice and style for your brand. \n","id":"collection-text-media#text-to-speech-4","customSortOrder":"4","heading":"How does text-to-speech work?"},"metadata":{"tags":[]}},{"fields":{"bodyContent":"

    Organizations implement text-to-speech technology in two ways. \n

    Self-managed \n

    AI/ML teams use specialized text-to-speech AI models and train them further with their own data. The model is then deployed to production and used across applications. The process is time-consuming and expensive. The organization is responsible for maintaining and managing the AI model. It can take months to get production-ready text-to-speech using this approach. \n

    Fully managed \n

    Fully-managed text-to-speech uses third-party models that you can integrate into code with APIs. A third-party provider fully manages, trains, and maintains the model. You give your text content to the model as input, and it generates audio files as output. You can also set it up to take web pages and other dynamically changing content as input and generate corresponding output in real time. \n

    Fully managed text-to-speech services are cost-effective and easy to use and integrate. You don't require ML/AI expertise to work with them. Developers can integrate these AI voice generators into existing applications in hours instead of days.","id":"collection-text-media#text-to-speech-5","customSortOrder":"5","heading":"How is text-to-speech technology implemented?"},"metadata":{"tags":[]}},{"fields":{"bodyContent":"

    Amazon Polly is a fully managed service that turns any text into lifelike speech. It is easy to use – you just send your text file to the Amazon Polly API, and it immediately returns the audio stream to play directly or store in a standard audio file format, such as MP3. Amazon Polly's pay-as-you-go pricing, low cost per request, and lack of restrictions on voice output reuse and storage make it a cost-effective way to enable speech synthesis everywhere. \n

    For example, with Amazon Polly, you can: \n

    \n
    • Convert text to speech in dozens of lifelike voices and languages to support all types of users. \n
    • Adjust the speech rate, pitch, or volume in output as needed. \n
    • Cache and replay generated speech at no additional cost. \n
    • Implement real-time text-to-speech capabilities at high speed and scale. \n \n

      You can also work with the Amazon Polly team to build a synthetic voice for your organization's exclusive use and differentiate your brand with a unique vocal identity. Amazon Polly is certified for use with regulated workloads for HIPAA (the Health Insurance Portability and Accountability Act of 1996) and Payment Card Industry Data Security Standard (PCI DSS). \n

      Get started with text-to-speech on AWS by creating a free account today.","id":"collection-text-media#text-to-speech-6","customSortOrder":"6","heading":"How can AWS support your text-to-speech projects?"},"metadata":{"tags":[]}}]},"metadata":{"auth":{},"pagination":{"empty":false,"present":true},"testAttributes":{}},"context":{"page":{"pageUrl":"https://aws.amazon.com/polly/what-is-text-to-speech/"},"environment":{"stage":"prod","region":"us-east-1"},"sdkVersion":"1.0.129"},"refMap":{"manifest.js":"47864e0370","rt-text-media-collection.rtl.css":"4a56177a31","rt-text-media-collection.js":"549552b6f3","rt-text-media-collection.css.js":"b7f4f2a16e","rt-text-media-collection.css":"dbf6370bb5","rt-text-media-collection.rtl.css.js":"641e7324ee"},"settings":{"templateMappings":{"hyperlinkText":"hyperlinkText","hyperlinkUrl":"hyperlinkUrl","heading":"heading","mediaAltText":"mediaAltText","mediaPosition":"mediaPosition","mediaUrl":"mediaUrl","subheader":"subheader","bodyContent":"bodyContent","videoOverlayTitle":"videoOverlayTitle","videoThumbnailUrl":"videoThumbnailUrl","videoPlayButtonText":"videoPlayButtonText","dark":"dark"}}}

      What is text-to-speech?

      Text-to-speech, also known as TTS, is a technology that converts written words into audible speech. An AI voice generator communicates with users when reading a screen is impossible or inconvenient. Text-to-speech technology opens up applications and information to be used in new ways, improving accessibility for individuals who cannot read text on a screen.

      Text-to-speech technology has evolved over the last few decades. Deep learning makes it possible to produce very natural-sounding speech that includes pitch, rate, pronunciation, and inflection changes. Today, computer-generated speech is used in various use cases and is becoming ubiquitous in user interfaces. Newsreaders, gaming, public announcement systems, e-learning, telephony, IoT apps and devices, and personal assistants are just starting points.

      What are the benefits of text-to-speech?

      Speech synthesis makes applications more accessible, allowing users to consume and comprehend information without having to focus on a screen. Here is a quick overview of some key advantages to using text-to-speech technology.

      Accessibility

      Text-to-speech caters to various communication styles and preferences, making digital content accessible to a broader audience. It improves access for users who cannot read due to visual impairments, literacy challenges, age, or other health concerns. As an assistive technology, it offers an alternative way to get information and ensure inclusivity.

      Enhanced learning

      Text-to-speech is applied to online materials to facilitate e-learning. Combining visual and audio presentations improves comprehension, recall, vocabulary skills, motivation, and confidence. The technology reads digital text aloud so language learners can understand how to pronounce words and phrases accurately. Hearing the text also reinforces vocabulary retention and sentence structure understanding.

      Mobility & freedom

      Text-to-speech can turn any digital content into a multimedia experience. People can listen to news, blog articles, or even a PDF document on the go or while multitasking. Flexibility boosts productivity as users consume content hands-free.

      Engagement and user experience

      TTS technology encourages users to engage with lengthy articles, reports, or books. They can access more written content in less time, improving content retention. It improves application metrics like visitor count and time spent on site. You get more conversions by enhancing the customer journey.

      Fast and affordable

      Cloud computing has made it fast and easy to implement text-to-speech. The cloud's economics of scale also make it inexpensive to integrate. You don't have to pay upfront or minimum monthly fees to start. You only pay if and when users access the feature.

      What are the use cases of text-to-speech technology?

      Applications that use voice to communicate are becoming more common every day. With text-to-speech solutions, your websites, mobile apps, digital books, e-learning tools, and online documents can literally have their own voice. We give some example use cases below.

      Audio publishing

      Publishers and content owners can quickly and inexpensively convert books, articles, and written material into audio with text-to-speech. You can convert existing written text to target a broad learner base for e-learning and training use cases. Turn your content into a more effective and less costly format to roll out across multiple languages.

      Customer service

      TTS systems enhance the quality of interactive call centers and support communication applications. Build better chatbots and AI assistants that read aloud digital text for users when requested. It is also a key technology in interactive response mechanisms and automated phone systems. Extend automated customer service interactions beyond monotonous phrases to conversational responses that feel empathetic and improve customer satisfaction.

      Media & entertainment

      TTS technology can be used to generate voiceovers for videos, animations, and interactive games. It lowers costs and increases efficiency in media pre-production and development. It also allows for real-time narration and dynamic commentary based on player actions in gaming or interactive apps. You can also use text-to-speech tools to deliver immersive audio content in virtual reality (VR) environments.

      Healthcare

      TTS technology in healthcare opens communication lines with patients and addresses the shortage of healthcare professionals. Generative AI-powered applications with voice interfaces can interpret patient queries and intent, triage patients, and respond in a natural-sounding voice. They can do everything from booking appointments to supporting treatment management and medicine reminders without forcing the patient to read a screen.

      How does text-to-speech work?

      Text-to-speech systems use powerful artificial intelligence (AI) and machine learning (ML) models to generate spoken words from text. The models run on deep neural networks—computing nodes that link and work together, like the human brain. The deep neural networks are trained on voice data in various languages, accents, pitch, and volume. During training, both the audio clip and its corresponding transcribed text are given to the AI model. The model identifies co-relations and patterns between the written and spoken text. It uses that knowledge to analyze and convert new text to sound.

      The process works as follows.

      Transforming text into time-aligned features

      The neural network first takes the input text and converts it into time-aligned features that represent the detailed characteristics of speech over time, such as pitch, rhythm, and tone. Common features include:

      • Mel spectrogram, which shows how sound frequencies change over time.
      • F0 frequencies, which represent the pitch or fundamental speech frequency.

      The system also considers linguistic features like how certain sounds should be pronounced or stressed, aligning them with the timing needed to make the speech sound natural. For example, the word hello has a short first sound followed by a long second sound.

      Converting time-aligned features into audio

      The next step transforms these features into audio that sounds like human speech. The neural network processes the features to synthesize a smooth and natural-sounding voice. Advanced text-to-speech technologies give you features like:

      • Volume control with whisper capability
      • High and low-pitch
      • Fast or slow speed.
      • Multiple languages and accents
      • Multiple speaking styles, including customized voice and style for your brand.

      How is text-to-speech technology implemented?

      Organizations implement text-to-speech technology in two ways.

      Self-managed

      AI/ML teams use specialized text-to-speech AI models and train them further with their own data. The model is then deployed to production and used across applications. The process is time-consuming and expensive. The organization is responsible for maintaining and managing the AI model. It can take months to get production-ready text-to-speech using this approach.

      Fully managed

      Fully-managed text-to-speech uses third-party models that you can integrate into code with APIs. A third-party provider fully manages, trains, and maintains the model. You give your text content to the model as input, and it generates audio files as output. You can also set it up to take web pages and other dynamically changing content as input and generate corresponding output in real time.

      Fully managed text-to-speech services are cost-effective and easy to use and integrate. You don't require ML/AI expertise to work with them. Developers can integrate these AI voice generators into existing applications in hours instead of days.

      How can AWS support your text-to-speech projects?

      Amazon Polly is a fully managed service that turns any text into lifelike speech. It is easy to use – you just send your text file to the Amazon Polly API, and it immediately returns the audio stream to play directly or store in a standard audio file format, such as MP3. Amazon Polly's pay-as-you-go pricing, low cost per request, and lack of restrictions on voice output reuse and storage make it a cost-effective way to enable speech synthesis everywhere.

      For example, with Amazon Polly, you can:

      • Convert text to speech in dozens of lifelike voices and languages to support all types of users.
      • Adjust the speech rate, pitch, or volume in output as needed.
      • Cache and replay generated speech at no additional cost.
      • Implement real-time text-to-speech capabilities at high speed and scale.

      You can also work with the Amazon Polly team to build a synthetic voice for your organization's exclusive use and differentiate your brand with a unique vocal identity. Amazon Polly is certified for use with regulated workloads for HIPAA (the Health Insurance Portability and Accountability Act of 1996) and Payment Card Industry Data Security Standard (PCI DSS).

      Get started with text-to-speech on AWS by creating a free account today.