Deepgram
Deepgram provides speech-to-text and text-to-speech APIs for developers, including Nova transcription models and Flux models built for voice agents.
About this tool
Deepgram is a voice AI platform used by developers to transcribe and generate speech. Its Nova models handle speech recognition in more than 45 languages with speaker diarisation, smart formatting, key-term prompting and automatic language detection, in real time or in batch. A redaction feature can remove sensitive personal data such as card or phone numbers from transcripts.
Its newer Flux models target voice agents. Flux speech-to-text is designed to understand turn-taking and interruptions, and Flux text-to-speech keeps tone consistent across a conversation. Deepgram can be used through its cloud or self-hosted for teams with data residency needs. New accounts receive free credit, and pricing is charged per minute or per thousand characters.
What you can do
Nova speech-to-text
Real-time and batch transcription in 45+ languages with diarisation, formatting and language detection.
Flux for voice agents
Conversation-aware models that handle turn-taking and interruptions more naturally in live dialogue.
Text-to-speech
Voices generated for live conversations, billed per thousand characters.
PII redaction
Sensitive details like card numbers and phone numbers can be removed automatically from transcripts.
Who is it for?
- A contact-centre platform transcribes calls in real time for agent assistance.
- A startup builds a voice agent that handles interruptions without awkward pauses.
- A healthcare company self-hosts transcription to keep audio within its own environment.
Before you choose
Per-minute rates differ by model, feature and plan, and promotional prices may change; check the current table before estimating costs.
Self-hosting is available for data control, but it adds infrastructure work compared with the cloud API.
Frequently asked questions
Does Deepgram have a free tier?
New users receive a $200 free credit with no minimums or credit card required, then pay as they go.
What is Deepgram Flux?
Flux is Deepgram’s family of conversation-aware speech models for voice agents, designed to manage turn-taking, interruptions and context.
Can Deepgram be self-hosted?
Yes. Deepgram states its models are available in the cloud or self-hosted, which helps teams with strict data requirements.
Check the official website for current features, pricing, and terms.
Similar tools
AssemblyAI
AudioAssemblyAI offers speech-to-text and voice AI APIs, including its Universal speech models, streaming transcription, speech understanding and a voice agent API.
Cartesia
AudioCartesia offers real-time speech generation, transcription and voice agent models built on state space models, accessed through a single API.
Vapi
AutomationVapi is a developer platform for building voice AI agents that handle phone and web calls, combining speech recognition, an LLM and speech synthesis.
Speechmatics
AudioSpeechmatics offers automated conversion of spoken media into written form, supporting transcription, translation, and language detection across audio, video, and text inputs.
FLUX by Black Forest Labs
Image & DesignFLUX is a family of image and video generation models from Black Forest Labs, available through an API, a playground and open weights for self-hosting.
FakeYou AI
AudioFakeYou AI offers a voice‑cloning service that turns written text, existing recordings, or selected languages into synthetic speech.