Deepgram

Deepgram is a developer-focused voice AI platform founded in 2015. Its APIs support speech recognition, speech synthesis, voice agents, and audio intelligence workflows.

The platform handles both streaming and prerecorded audio. Developers can access services through REST APIs, WebSocket connections, official SDKs, a command-line interface, and hosted or self-hosted deployments.

Common applications include contact-center transcription, voice assistants, conversational AI, call analytics, meeting transcription, media indexing, and domain-specific speech processing.

Deepgram’s speech-to-text services transcribe streaming, prerecorded, and turn-based audio. Available capabilities include model selection, language support, speaker separation, formatting, and audio analysis options.

The platform also provides text-to-speech APIs, a Voice Agent API, and Audio Intelligence features. Voice-agent workflows can combine speech recognition, language models, turn handling, and speech synthesis.

Official SDKs are available for Python, JavaScript or TypeScript, Go, .NET, Java, and Rust. Deepgram also provides a CLI, API reference, Playground, community resources, and documentation.

Supported integration patterns include Twilio, Amazon Connect, Amazon SageMaker, Amazon Bedrock, LiveKit, Daily, Vapi, Five9, Genesys, Cognigy, Vonage, and other partner platforms. Availability and implementation details vary by integration.

Deepgram uses usage-based pricing for its public APIs. New accounts receive $200 in free credit, and the pricing page states that no credit card is required for the free starting option.

The Pay As You Go option has no minimum commitment and no credit expiration. The Growth option begins at $4,000 or more per year and offers volume-based savings. Enterprise pricing is customized.

Deepgram’s pricing page lists Nova-3 speech-to-text at $0.29 per hour for monolingual streaming and $0.35 per hour for multilingual streaming. Flux is listed at $0.39 per hour for monolingual use and $0.47 per hour for multilingual use. Rates vary by model, endpoint, and plan.

Users can test the APIs through the Playground without writing code. The free account and credit function as the available trial mechanism rather than a separate time-limited subscription.

Pros

• Provides speech-to-text, text-to-speech, voice-agent, and audio-intelligence APIs.
• Supports both real-time streaming and prerecorded audio processing.
• Offers official SDKs across several major programming languages.
• Includes cloud, regional, dedicated, and self-hosted deployment options.
• Provides free credits and usage-based pricing without a stated minimum commitment.

Cons

• The product is primarily API-oriented and requires development work for most applications.
• Total costs depend on audio volume, model selection, concurrency, and endpoint type.
• Enterprise deployment, dedicated infrastructure, and custom requirements may require sales engagement.
• Mobile and desktop users do not receive a conventional standalone transcription application.

Comparable services include AssemblyAI, Google Cloud Speech-to-Text, Microsoft Azure AI Speech, Amazon Transcribe, and Speechmatics. These products also provide developer APIs for speech recognition, transcription, or related voice workflows.

Deepgram is most directly comparable to AssemblyAI and Speechmatics for developer speech APIs. Google Cloud, Microsoft Azure, and Amazon Web Services may be preferable for organizations standardizing on their broader cloud ecosystems.

Does Deepgram offer a free trial?

Yes. New accounts include $200 in free credit. Deepgram also provides a Playground for testing API requests, and its pricing page states that no credit card is required for the free starting option.

Does Deepgram have an API?

Yes. Deepgram provides REST and WebSocket APIs for speech-to-text, text-to-speech, voice agents, audio intelligence, model discovery, account administration, and related functions.

Which programming languages does Deepgram support?

Official SDKs are available for Python, JavaScript or TypeScript, Go, .NET, Java, and Rust. Developers can also use cURL and other HTTPS clients.

Can Deepgram be self-hosted?

Yes. Deepgram documents self-hosted, dedicated, regional, and cloud deployment options. Regional endpoints include European Union and Australian processing endpoints.

Which integrations are available?

Documented or partner-supported integrations include Twilio, Amazon Connect, Amazon SageMaker, Amazon Bedrock, LiveKit, Daily, Vapi, Five9, Genesys, Cognigy, and Vonage.

Is Deepgram available for Windows and macOS?

The Deepgram CLI supports Windows, macOS, and Linux. The main product is accessed through web tools, APIs, SDKs, and deployment environments rather than native consumer desktop applications.