Dictation

Power real-time dictation with fast, accurate Voice AI

Give professionals instant, accurate voice-to-text input that keeps pace with natural speech across clinical, legal, and enterprise workflows.

Get started

increase in free-to-paid conversion rate

See case study

“The new Universal-3.5 Pro speech model from AssemblyAI is best so far in terms of accuracy, latency, and language switching.”

80%

increase in customer satisfaction

See case study

“Assembly has saved us countless hours managing models, and provided exceptional accuracy.”

increase in free-to-paid conversion rate

See case study

“The new Universal-3.5 Pro speech model from AssemblyAI is best so far in terms of accuracy, latency, and language switching.”

80%

increase in customer satisfaction

See case study

“Assembly has saved us countless hours managing models, and provided exceptional accuracy.”

What is a dictation API?

A dictation API turns a single spoken utterance into finished, formatted text in one call. It transcribes what was said, then applies an LLM cleanup pass that resolves self-corrections, drops filler, and returns text ready to go straight into whatever your user is writing.

How voice-to-text input actually works

Capabilities

Built for every dictation workflow

Enable hands-free documentation with real-time voice-to-text that professionals can trust.

Latency that keeps pace with natural speech

Dictation API: finished, formatted text back in under a second

Sync API: a verbatim transcript in one call, ~134 ms at p50

Realtime API: sub-300 ms for continuous live dictation

Domain-specific accuracy across professions

Keyterms prompting for product names, customer names, and internal jargon

Prompting that tells the model what kind of speech to expect

Medical Mode for clinical terminology, on pre-recorded and realtime transcription

Price-performance and scalability that grows with you

Pay as you go with no minimum commitment

Unlock volume discounts as you scale

Count on 99.9% uptime SLA at a fraction of legacy pricing

Platform

One API, every dictation workflow

Real-time accuracy that keeps pace with natural speech, with domain-specific precision across clinical, legal, and enterprise terminology.

Dictation API

The first API built for dictation: flagship transcription plus an LLM cleanup pass in a single call, so speech comes back already formatted by the instructions you send.

Learn more Try for free

Sync Speech-to-Text API

Speak, see it typed. One HTTP call returns a finished, flagship-accuracy transcript in ~134 ms — built for push-to-talk dictation, voice input fields, and short clips up to 2 minutes.

See the models Try for free

Realtime Speech-to-Text API

Low-latency real-time accuracy for continuous live dictation across clinical, legal, and enterprise workflows. Immediate text output with automatic punctuation, casing, and formatting.

See the models Try for free

Which request shape does your feature need

Four ways to get speech into text, separated by what the user is waiting for. The row that usually decides it is when the text appears.

What you need

Dictation API Finished text

Sync API Verbatim transcript

Realtime API Live transcript

Pre-recorded Async job

What comes back

Cleaned-up text, plus the verbatim transcript alongside it

A verbatim transcript

A transcript that builds as they speak

A transcript once the job finishes

When text appears

After key release

Mid-sentence, while they are still talking

Later, once processing completes

Audio shape

One utterance, up to 120 seconds

One clip, up to 2 minutes

An open-ended live stream

A complete file

Transport

One HTTP POST

A WebSocket

Submit, then poll or take a webhook

Who is waiting

The user, mid-task

The user, continuously

Nobody — it is a background job

LLM cleanup in the same call

Included

—

Reach for it when

The text goes straight into whatever your user is writing

You want the transcript itself, not a rewrite

A live session needs text as it happens

You have a recording and time to process it

Dictation library

Which page answers which question

Everything we have published on dictation, and the question each piece answers.

Where to look	The question it answers
What is a dictation API? Blog	What a dictation API is, and how it differs from plain speech-to-text.
Inside dictation cleanup Blog	How raw speech becomes finished, formatted text.
Dictation features Blog	What product teams are shipping in 2026.
Introducing the Dictation API Announcement	What the Dictation API does, and what shipped at launch.
Sync API or Dictation API? Blog	Which endpoint your feature needs: the transcript, or the finished text.
Dictation API vs speech-to-text + an LLM Blog	Whether to buy the bundle or build it from two vendors.
How to evaluate a dictation API Blog	How to test dictation APIs before committing to one.
Best dictation APIs for developers Blog	Which dictation APIs exist, and how they line up against each other.
Voice coding Blog	How developers dictate to Claude Code, Cursor, and Copilot.
Build a dictation app with the Sync API Tutorial	A working dictation app, end to end, on the Sync API.
How to build push-to-talk dictation Tutorial	How to wire push-to-talk capture to a transcription call.
How to build clinical dictation Tutorial	How to build clinical dictation, and where Medical Mode runs.
Blurt Open source	A free, open-source macOS dictation app built on the API.
Dictation API Product	Price, languages, latency, and how it compares to Sync and Realtime.
Dictation API documentation Docs	The endpoint, config fields, audio limits, and error handling.

Benchmarks

Lowest error rate in the industry

The crucial details captured more accurately than ever — powering reliable downstream workflows.

	AssemblyAI Universal-3.5 Pro	Speechmatics Enhanced Model	Deepgram Nova-3	AWS Transcribe	OpenAI GPT-4o Transcribe
MER	7.5%	17.33%	18.69%	20.76%	12.29%
WER	4.5%	6.1%	6.66%	12.9%	5.34%

Playground

We're not playing around, but you can

Put our Voice AI models to the test in our no-code playground.

Try it now