For AI agents: a documentation index is available at /llms.txt. Markdown versions of all pages can be requested by appending `.md` to the URL, or by setting the `Accept` header to `text/markdown`.
Skip to main content
Speech to Text

Speech to Text overview

Learn how to turn audio into text.

Use Speech to Text to transcribe using one of the modes:

  • Realtime processing: Stream audio from an input device or file and receive instant updates of the transcription as it happens
  • Batch processing: Submit an audio file and receive a complete text transcription once the processing is finished
  • Agent STT: Stream audio from a conversation and receive speaker-labelled, turn-based transcription for a voice agent

Developer quickstart​

Transcribe in real time

Instantly convert streaming audio to text with Realtime processing

Transcribe a file

Submit an audio file and receive a complete text transcription once the processing is finished

Transcribe for voice agents

Stream audio and receive speaker-labelled, turn-based transcription ready to pass to an LLM

The quickest way to transcribe voice from audio is in our web portal.

Deployments​

Speechmatics provides flexible deployment options tailored to your requirements. You can host the platform in your own environment, use our managed service, or choose a hybrid approach.

For deploying our API in your environment, contact sales, or see our on-prem documentation.

Realtime processing​

Turn live audio into accurate transcripts — instantly.

The Speechmatics Realtime Speech to Text API converts spoken audio into text with low latency and high accuracy.

  • Use when speed matters
  • Transcribe live broadcasts or events
  • Caption webinars, meetings, or podcasts in real time
  • Power voice assistants or AI agents with live input
  • Monitor contact center calls as they happen
  • Build accessibility features like live captions

Batch processing​

Create transcripts from pre-recorded audio or video.

The Speechmatics Batch Speech to Text API processes pre-recorded files asynchronously by default, returning highly accurate transcripts in a range of formats. To block for the result in a single request instead, use synchronous transcription.

  • Transcribe recorded meetings or interviews
  • Caption on-demand videos and podcasts
  • Generate searchable transcripts for media archives
  • Process customer service recordings for compliance or insights
  • Automate subtitles across large video libraries

What is a job?​

Each transcription request is handled as a job — a self-contained unit representing a single transcription task.

A job includes:

  • The audio or video file to be transcribed
  • Configuration settings (e.g. language, formatting, diarization)
  • Metadata and status tracking
  • The resulting transcript(s)

You submit a job to the API, monitor its progress, and retrieve results once it's complete. Jobs can be created via direct upload or by referencing a URL.

Agent STT​

Transcription built for voice agents and conversational applications.

Agent STT groups the transcript into segments: punctuated, attributed to a speaker, and ready to pass to an LLM. It also marks where each speaker's turn starts and ends, so your agent knows when to respond.

  • Build a voice agent or live conversational assistant
  • Respond as soon as a caller stops speaking
  • Feed clean, speaker-labelled turns to an LLM
  • Handle multi-speaker calls without stitching words together

Languages

See the languages Speechmatics supports

Pricing

Compare our pricing options

Translation

Translate your transcript to another language

Custom dictionary

Add custom words to your transcript

Diarization

Identify speakers in your transcript

Accuracy benchmarking

Benchmark the accuracy of our models