AI & Machine Learningdeepgram.com &nearr;

Deepgram API for AI Agents

Speech-to-text and audio intelligence

Deepgram provides APIs for speech recognition, transcription, and audio analysis. AI agents can use Deepgram to transcribe audio, detect speakers, analyze sentiment, and process real-time speech.

What AI agents can do with Deepgram

Structured actions an AI agent can execute through the Deepgram API

Action

Description

Inputs

Outputs

transcribeAudio

Transcribe audio to text

audio_url, model, language, punctuate, diarize

transcript, words[], confidence

transcribeStream

Real-time streaming transcription

audio_stream, model, interim_results

transcript, is_final, speech_final

analyzeText

Analyze text for topics and sentiment

text, language, topics, sentiment

topics[], sentiment, summary

Use cases for Deepgram + AI agents

Meeting transcription and summarization
Real-time captioning
Voice command processing for agents
Call center analytics
Podcast and video transcription

How to connect Deepgram to an AI agent

1Get your Deepgram API key
2Generate an AgentSpec for transcription actions
3Define batch and streaming transcription actions
4Publish for discovery
5Test with sample audio files

Best practices

✓Enable speaker diarization for multi-speaker audio

✓Use the appropriate model for your domain (general, meeting, phone)

✓Enable punctuation and smart formatting

✓Handle streaming results with interim vs final transcript

✓Choose language model based on audio quality and content

Frequently asked questions

How do AI agents transcribe audio with Deepgram?+

Send an audio URL or audio data to the transcribeAudio action. Deepgram returns a transcript with word-level timestamps, confidence scores, and optional speaker labels.

Can agents do real-time transcription?+

Yes, using the streaming API. Send audio chunks and receive interim and final transcripts in real-time. This enables live captioning and voice command processing.

What languages does Deepgram support?+

Over 30 languages including English, Spanish, French, German, Portuguese, Japanese, Korean, and more. Check the Deepgram docs for the full list and model availability per language.