Deepgram STT/TTS for Voice AI: Speech Capabilities, APIs, Integrations & Implementation
Deepgram provides a streaming-first speech stack: real-time and batch STT (Flux, Nova), Flux/Aura TTS, Voice Agent orchestration, SDKs, callbacks, and documented integrations for contact-center and telephony platforms.
Peak Demand evaluates the platform in the context of telephony, APIs, business rules, integrations, QA, monitoring, and the operating environment around the agent.
Discuss a Deepgram STT/TTS DeploymentWhat Is Deepgram STT/TTS?
Deepgram is a speech component platform for enterprise STT and TTS with streaming-first WebSocket APIs, low-latency Flux/Nova STT and Flux/Aura TTS, Voice Agent orchestration, SDKs, callbacks, and official integration guides for contact-center and telephony partners.
Deepgram STT/TTS Platform Profile
Deepgram STT/TTS
Speech component • STT/TTS
Where Deepgram STT/TTS Fits in a Voice AI Technology Stack
Core component — Speech component (STT/TTS) used as the speech layer in real-time voice agents, contact-center analytics, live agent assist, or media transcription pipelines.
A Typical Deepgram STT/TTS Production Architecture
The exact architecture depends on the business environment, but Peak Demand evaluates the platform as one layer inside a connected production system.
Deepgram STT/TTSSpeech component / STT/TTSDeepgram STT/TTS Capabilities Relevant to Production Voice AI
| Capability | Current position | Scope | Implementation context |
|---|---|---|---|
| Inbound calling | Limited / conditional | External integration | Inbound phone calls are handled by integrating Deepgram with telephony/contact-center providers (examples and guides exist for Amazon Connect, Twilio and other integrations); Deepgram processes audio from those providers rather than acting as a PSTN carrier itself. |
| Outbound calling | Limited / conditional | External integration | Outbound call capabilities are achieved by integrating Deepgram with telephony providers (e.g., Twilio, Amazon Connect) that originate calls; Deepgram provides STT/TTS and agent orchestration on the audio stream. |
| Telephony / phone routing | Limited / conditional | External integration | Deepgram documents integrations and examples for contact-center routing and CRM pipelines (e.g., Amazon Connect, contact-center analytics) but phone routing itself is implemented via partner platforms; Deepgram provides transcription/analytics to enable intelligent routing. |
| SIP / trunking | Limited / conditional | External integration | Deepgram documents integrations with SIP-capable appliances and hubs (AudioCodes LiveHub is cataloged in integrations), indicating SIP/trunking capture is achieved via partner hardware/services rather than built into Deepgram as a carrier. |
| Webhooks / callbacks | Established | Product-native | Deepgram supports STT and TTS callback/webhook workflows. STT Callback and TTS Callback let you supply URLs for asynchronous results; streaming callbacks support http/https/ws/wss and include authentication options (Basic Auth, dg-token header). |
| Public APIs | Established | Product-native | Deepgram exposes documented public APIs covering Speech-to-Text, Text-to-Speech, Voice Agent, Audio Intelligence and account/billing management; API overview and references are published in developer docs. |
| SDKs / developer libraries | Established | Product-native | Official SDKs and language bindings are published (JavaScript, Python, .NET, Go, Java) and referenced in developer docs and guides. |
| Tool / function calls | Established | Platform-family | Voice Agent API includes orchestration primitives such as function calling and mid-session control to enable action-taking and integrations during conversations; function-calling is part of the Voice Agent platform-level features, not merged into raw STT/TTS primitives. |
| Transfers / forwarding / handoff | Limited / conditional | Platform-family | Deepgram’s streaming transcription and Voice Agent tooling enable intelligent routing and handoff decisions (e.g., routing dissatisfied callers to specialists) but actual call transfer/forwarding is handled via telephony/contact-center integrations. |
| Conference / queue primitives | Not found in reviewed official docs | Not applicable / unresolved | Not found in the official documentation reviewed for this profile; this is not a claim that the capability is unsupported. |
| Appointment booking | Not found in reviewed official docs | External integration | Not found in the official documentation reviewed for this profile; this is not a claim that the capability is unsupported. |
| Calendar integration | Not found in reviewed official docs | External integration | Not found in the official documentation reviewed for this profile; this is not a claim that the capability is unsupported. |
| Knowledge bases / retrieval | Limited / conditional | Platform-family | Deepgram provides Audio Intelligence (summarization, intent, topic detection) and the Voice Agent API supports BYO LLM orchestration; retrieval-style KB integration requires connecting an external LLM/KB or using BYO LLM via the Voice Agent/platform orchestration. |
| Workflow automation | Limited / conditional | Platform-family | Deepgram offers orchestration primitives in the Voice Agent API plus callbacks/webhooks and published integrations (Zapier/Make) to implement automated workflows; full workflow engines are external or implemented via platform-level orchestration. |
| Integrations / connectors | Established | Platform-family | Developer docs contain integration guides for many platforms and partners (Amazon Connect, Twilio, LiveKit, Dialogflow CX, AudioCodes LiveHub, Zapier, Zoom, etc.), showing official integration patterns and examples. |
| Call recording | Limited / conditional | Product-native | Deepgram supports processing pre-recorded audio (batch transcription) and asynchronous callbacks for results; capturing/hosting raw PSTN call recordings is performed by the telephony/capture layer (partner platforms), while Deepgram provides ingestion and transcription of those recordings. |
| Transcription / speech-to-text | Established | Product-native | Deepgram’s Speech-to-Text API (Flux STT, Nova models) provides real-time and batch transcription, model options for conversational turn detection, diarization, punctuation, redaction, and language coverage (50+ languages). |
| Text-to-speech / voices | Established | Product-native | Deepgram offers Flux TTS (conversation-native, very low latency, expressive) and Aura-2 (multi-language) with streaming (WebSocket) and batch REST APIs; voices cataloged with model strings and voice characteristics. |
| Realtime audio / media streaming | Established | Product-native | Streaming WebSocket transports are documented for real-time STT (Flux/Nova) and streaming TTS (Flux/Aura). Deepgram advertises ultra-low latency (sub-300ms STT, TTS start as low as 80ms) for real-time voice agents. |
| DTMF / speech gather | Not found in reviewed official docs | Not applicable / unresolved | Not found in the official documentation reviewed for this profile; this is not a claim that the capability is unsupported. |
| Call logs / analytics / observability | Established | Platform-family | Deepgram provides speech analytics capabilities (Audio Intelligence) including summarization, sentiment, intent detection, topic detection, KPIs and call analytics; docs also reference logs & usage data and billing/usage endpoints in the API overview. |
| Testing / simulation | Established | Product-native | Deepgram provides an API Playground and interactive demos for testing STT/TTS/Voice Agent flows; developer docs include starter kits and sample apps for simulation and testing. |
| Language support | Established | Product-native | STT supports 50+ languages and multiple models (Flux STT for conversational applications in 10 languages; Nova for 50+ languages). Aura-2 TTS supports multiple languages (English, Spanish, German, French, Dutch, Italian, Japanese); Flux TTS currently focused on English voices. |
| Security / compliance | Established | Platform-family | Deepgram documents data privacy and compliance policies; platform options include HIPAA support, GDPR, regional data residency, dedicated runtimes/self-hosting and guidance on callback authentication mechanisms. |
| Pricing / billing model | Established | Platform-family | Deepgram documents pricing for components: Flux/Aura TTS pricing (example: $0.045 per 1,000 characters for Aura/Flux TTS listing) and Voice Agent API pricing (flat-rate $4.50/hr). The platform also exposes billing and usage endpoints in the API overview. |
Capabilities marked “Not found in reviewed official docs” were not located in the official documentation corpus reviewed for this profile; that status does not mean the capability is unsupported.
How Deepgram STT/TTS Can Connect to Business Systems
Deepgram provides published integration guides and examples for Amazon Connect, Twilio, LiveKit, AudioCodes LiveHub, Dialogflow CX, Zapier, Zoom and others. Telephony capture, PSTN origination/termination, SIP/trunking and call routing are generally achieved via these partner platforms; Deepgram consumes and processes the audio streams supplied by them.
Official integration guides exist for Amazon Connect, Twilio, LiveKit, AudioCodes LiveHub, Dialogflow CX, Zapier, Zoom and others (see integrations docs).
Telephony capture, SIP/trunking, and PSTN call origination/termination are handled via partner platforms; Deepgram processes audio streams from those integrations.
Common Deepgram STT/TTS Use Cases
Voice-level orchestration is available via the Voice Agent API (function-calling, mid-session control) plus webhooks/callbacks and Zapier/Make examples. Use callbacks and the Voice Agent primitives to build automated routing, CRM pipelines, and orchestration, while relying on your telephony provider for PSTN-level actions.
Real-time conversational AI agents (listen → think → speak) using Flux STT and Flux/Aura TTS
Streaming transcription for speech analytics, QA, and real-time agent assist
Batch transcription and captioning for media/podcasts with high accuracy across noisy audio
Where Deepgram STT/TTS May Be Particularly Strong
Documented strengths include ultra low-latency conversational STT (Flux/Nova) and low-latency Flux/Aura TTS for real-time agents, a unified STT→agent→TTS stack, streaming-first APIs with WebSocket transports, and flexible enterprise deployment and compliance options.
Strength 1
Ultra low-latency conversational STT (Flux / Nova) and TTS (Flux TTS / Aura) suitable for real-time agents
Strength 2
Unified voice stack (STT + LLM orchestration + TTS) and voice-agent orchestration
Strength 3
Streaming-first APIs (WebSocket) with callbacks and SDKs for major languages
Strength 4
Flexible deployment: cloud, VPC, self-hosted; enterprise compliance (HIPAA/GDPR) options
Where Deepgram STT/TTS May Not Be the Best Fit
Deepgram is integration-first for telephony: PSTN/SIP/phone-routing primitives are implemented via partner platforms rather than carrier-native features. Flux TTS is conversation-native today but English-focused; Aura-2 covers additional languages until Flux multilingual voices ship. Full voice-agent orchestration features are in the broader Voice Agent API rather than the raw STT/TTS endpoints.
Consideration 1
Telephony primitives (PSTN, SIP/trunking, routing, queues) are achieved via documented integrations rather than being carrier-native
Consideration 2
Flux TTS is conversation-native but currently English-first (other languages served by Aura-2 until Flux multilingual ships)
Consideration 3
Full voice-agent orchestration features live in the wider Voice Agent API/platform rather than the raw STT/TTS primitives
When Peak Demand May Choose Deepgram STT/TTS
Choose Deepgram when you need a low-latency, streaming-first STT/TTS layer and voice-agent orchestration with enterprise deployment/compliance options. If you require built-in carrier services (phone-number provisioning, native call routing/queues) without a telephony partner, evaluate how those needs will be fulfilled by a telephony provider alongside Deepgram.
Best-fit pattern 1
Real-time voice agents and IVR built around low-latency STT and expressive TTS
Best-fit pattern 2
Contact-center analytics and live agent assist using streaming transcription and Audio Intelligence
Best-fit pattern 3
Enterprises requiring deployment flexibility and compliance (self-hosted / VPC / HIPAA)
When another platform may deserve a closer look
Evaluate alternatives when 1
Standalone PSTN carrier services (phone-number provisioning, native call-routing/queues) without a telephony provider
Evaluate alternatives when 2
Use cases requiring built-in telephony SBC/SIP trunk features without integration to a telephony partner
Security, Data Handling & Compliance Considerations
Deepgram documents data privacy and compliance options (including HIPAA and GDPR references), regional data residency and dedicated runtimes/self-hosting. Callback/webhook requests support authentication mechanisms such as Basic Auth or dg-token headers; developer docs include guidance on retry/backoff behavior for failed callbacks.
Deepgram documents data privacy & compliance policies and enterprise deployment modes (self-hosted, VPC, on-prem) and explicitly references HIPAA/GDPR support and regional data residency options.
Callback/webhook requests support authentication via Basic Auth or dg-token header; retry/backoff behavior for failed callbacks is documented.
How Deepgram STT/TTS Pricing Should Be Evaluated
Deepgram publishes component pricing in product docs: example TTS pricing appears for Aura/Flux (example listing: $0.045 per 1,000 characters) and Voice Agent API pricing is published (example: $4.50/hr for the Deepgram full-stack rate). Billing and usage endpoints are exposed in the API for programmatic access.
Flux/Aura TTS example pricing shown on product pages (example: $0.045 per 1,000 characters) and promotional free-build trials for Flux TTS are documented.
Voice Agent API pricing is documented (flat-rate $4.50/hr for Deepgram’s full stack) and platform billing/usage endpoints are available in the API.
Billing and usage breakdown endpoints are exposed via the API overview for programmatic access to balances and billing data.
Testing the Platform Before Production
Developer docs and an API Playground provide interactive demos for STT, TTS and Voice Agent flows. Official SDKs, starter kits and sample apps (JavaScript, Python, .NET, Go, Java) are available for local simulation and integration testing.
What Peak Demand Adds Around Deepgram STT/TTS
Deepgram is streaming-first: real-time STT and TTS use WebSocket transports and REST endpoints for batch workflows. Production patterns combine: SDKs or raw WebSockets for low-latency audio, STT/TTS callbacks for asynchronous processing, Voice Agent API for orchestration and BYO-LLM integrations, and partner telephony platforms to handle PSTN capture/origination. Flux STT and Flux TTS target conversation-native low-latency flows; Aura-2 provides multi-language TTS voices. Verify telephony capture, supported transports and any integration limits with both Deepgram documentation and your chosen telephony provider.
Discovery & Platform Fit
Determine whether the platform is actually the right choice for the workflow before building around it.
Conversation & Agent Architecture
Design prompts, flows, variables, tools, validation, escalation and business logic.
Telephony & Realtime Infrastructure
Configure the appropriate phone, SIP, CPaaS or realtime transport layer for the deployment.
Middleware & APIs
Build controlled AWS, Cloudflare, API, webhook or middleware layers where systems require additional validation and orchestration.
Business-System Integration
Connect CRM, scheduling, EMR/EHR, ERP, databases, helpdesk, ordering, field-service or proprietary software where suitable integration surfaces exist.
QA & Managed Operations
Test workflows, monitor production behavior, review failures, measure outcomes and refine the implementation over time.
Deepgram STT/TTS Questions
Which languages and models does Deepgram support for STT and TTS?
Deepgram documents Flux STT for conversational applications in roughly 10 languages and Nova models covering 50+ languages for STT. For TTS, Flux TTS is conversation-native and currently English-focused; Aura-2 supports multiple languages including English, Spanish, German, French, Dutch, Italian and Japanese.
Can Deepgram originate or terminate PSTN calls directly?
Deepgram’s documentation shows integrations with telephony providers (Amazon Connect, Twilio, AudioCodes, etc.) for handling inbound and outbound calling. Deepgram processes audio streams supplied by those providers rather than acting as a PSTN carrier itself; telephony origination/termination is handled by the partner platform.
Does Deepgram offer low-latency real-time speech suitable for voice agents?
Yes. Deepgram documents streaming-first WebSocket transports for real-time STT (Flux/Nova) and streaming TTS (Flux/Aura). The platform advertises ultra-low latency characteristics (examples cited: sub-300ms STT and TTS starts as low as ~80ms) for real-time agent use cases.
How are callbacks, webhooks and authentication handled?
Deepgram supports STT and TTS callbacks/webhooks for asynchronous results; supported transports include http/https/ws/wss. Callback authentication options documented include Basic Auth and a dg-token header. Docs also cover retry and backoff behavior for failed callbacks.
Are conference/queue primitives or DTMF gather documented?
Conference/queue primitives and DTMF gather were not found in the official documentation reviewed for this profile. This notation reflects the reviewed docs and is not an assertion about product capability beyond what the documentation shows; for PSTN/telephony primitives, review Deepgram’s integration guides and your telephony provider’s features.
Can I use Deepgram with my CRM or automation platforms?
Yes. Developer docs include integration guides and examples for building CRM pipelines and automation using the Voice Agent API, webhooks/callbacks and published connectors such as Zapier. Orchestration primitives in the Voice Agent API enable mid-session control and function-calling for integrations.
What deployment and compliance options are available for enterprises?
Deepgram documents flexible deployment options including cloud, VPC and self-hosting/dedicated runtimes, and references HIPAA and GDPR compliance and regional data residency options. Review the security and trust docs for details and any enterprise contracting requirements.
Explore Retell AI
Retell AI is one of the full-stack Voice AI platforms Peak Demand evaluates for phone-first deployments, custom integrations, telephony, APIs, and managed production workflows.
Explore Retell AIPeak Demand may earn a commission from this link.
Planning a Deepgram STT/TTS Deployment?
Peak Demand can help evaluate platform fit, design the architecture, connect telephony and business systems, implement controlled tools and integrations, test edge cases, and manage the operational layer after launch.
Discuss a Deepgram STT/TTS DeploymentOfficial Deepgram STT/TTS Sources Reviewed
This profile is maintained using official or first-party vendor sources. Current vendor documentation remains the source of truth for an active production decision.
Last researched: 2026-08-16
Next recommended review: 2026-11-14
