What is the difference between speech to text and text to speech?

Speech to text turns spoken words into written text. Text to speech does the reverse: it reads written text aloud. Dictation software is speech to text. Some dictation products also include text to speech, so it is worth checking the feature list instead of assuming. Dragon is one example. The overview of dictation software and its guide to dictation software types explain how speech to text works in each kind of product.

Speech to text is what dictation does

When you dictate, the software recognizes your speech and types the words where the cursor is ¹. That is speech to text. It is also called automatic speech recognition, and the glossary entry gives the technical term. Accuracy is often reported as word error rate, a measure of how many words differ from a correct transcript.

Text to speech reads text aloud

Text to speech takes written text and speaks it in a computer voice. Some dictation products add it for review. Nuance's 2021 help for Dragon NaturallySpeaking says the Read That feature allows Dragon to read editable text aloud in a computer voice ². The Nuance help page notes that earlier Dragon versions called this feature Text-to-speech ³, so an older guide may use that name.

Read That is not the same as Playback

Dragon also has a Playback feature, and the two are easy to confuse. Nuance says that, unlike Playback, Dragon does not record your dictation and save it when it reads text ⁴. Read That speaks the text in the document. Playback replays what you dictated. Both are features of one product, so check which one you need.

On a Mac, reading text aloud is a separate setting

On a Mac, text to speech is a separate feature from Dictation. Apple says that once you turn on Speak selection, a keyboard shortcut makes your Mac start speaking, and the default is Option-Esc ⁵. You turn on Speak selection in the Read & Speak settings. Dictation, by contrast, enters your speech as text where the cursor is ¹. If you dictate on a Mac and want to hear text read back, check this setting, then test it on a sample document. Apple's steps may change between macOS versions, so use the menu names on your own screen.

What the text to speech feature reads

Read That works on text you can edit. Nuance says Dragon simply reads editable text, such as a Microsoft Word document or an email message ⁶. Use it on the finished document, not on a raw recording. Listening alone is not a reliable check: in one in-lab study, participants missed over 50 percent of speech recognition errors when they listened to the recognized text read aloud ⁷. Read the text on screen as well.

Text to speech can also imitate a voice

Text to speech appears in the security discussion too. The IEEE notes that speaker recognition systems can be attacked by synthesizing speech with text-to-speech tools ⁸. A voice that sounds right is therefore not proof of who is speaking. The question on speaker recognition covers that separate task.

Check the feature for your version

The Nuance page dates from a 2021 help release ², and commands and settings can change between versions. Before you rely on text to speech, confirm that your version has the feature and test it with a sample document. Listen for the words you dictated, and check that punctuation and numbers are read the way you expect. If you only need your dictated text read back to check it, a read-through by eye works too, and the question on how to write a dictation disclaimer covers review for accuracy.