Glossary
Generative AI
Generative 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.
Updated October 17, 20263 cited facts, each traced to its source
Also called GenAI, generative artificial intelligence.
Generative AI is the class of AI models that produce new content. The U.S. National Institute of Standards and Technology (NIST) defines it as models that emulate the structure and characteristics of input data in order to generate derived synthetic content 1. That content can be text, images, video, audio or other digital material. A chatbot's reply, a generated picture and a summary of your voice notes are all outputs of generative AI.
Generating is different from recognizing
A speech recognizer decides which words match a recording, so its output is a transcript of what was said. A generative model produces words that were not in its input. In dictation the difference matters. A plain recognizer writes what it heard. A generative model can rewrite it, which can fix punctuation and remove filler words, but can also change your meaning.
The term has no formal definition
Google's Machine Learning Glossary calls generative AI "An emerging transformative field with no formal definition" 2. Other documents give more precise working definitions, so the same phrase can mean slightly different things. When a product says it uses generative AI, ask what the model is asked to produce, and whether it adds words or only changes existing ones.
A concrete example: cleaning up a dictated note
Suppose you dictate: "um so the meeting is uh moved to thursday, I think three pm." A plain recognizer may return a transcript close to that wording, with the hesitations in it. A generative cleanup step might return "The meeting has moved to Thursday, I think at 3 p.m." The model supplied "has", "at" and "p.m.", which you did not say. That added wording is the generation, and it is useful only if the meaning is still yours.
Why it matters for dictation
Most dictation products that advertise AI add a generative step after the speech recognizer, so the choice decides where your words are processed. OpenAI says that when you press the microphone icon in ChatGPT, the recorded audio is sent to its models to be transcribed, and the returned text can be edited before you send it 3. Other products run the same steps on your device, and some offer both. Check the privacy page before you dictate sensitive material.
A rewrite step also costs time and may change what you meant. Our guide to AI cleanup for dictation covers the trade-off, and speech to text without AI explains how to get raw output instead.
Related terms
- Machine learning: the broad method that most generative models are built with.
- Large language model (LLM): the text generators behind most chatbots and writing tools.
- Prompt (AI): the input that tells a generative model what to produce.
- Generative error correction (GER): a generative step that rewrites speech recognizer output.
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.NIST's 2024 Generative AI Profile defines generative AI as the class of AI models that emulate the structure and characteristics of input data in order to generate derived synthetic content, which it says can include audio and text.
National Institute of Standards and Technology, National Institute of Standards and Technology, U.S. Department of Commerce, 2024 · Original
- 2.
- 3.
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 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.
- 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.
- 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.