# Deepgram Voice AI Platform > Deepgram is a foundational Voice AI platform for developers, offering unified APIs for Speech-to-Text (STT), Text-to-Speech (TTS), and autonomous Voice Agent orchestration. Available as a managed API (US, EU, and Australia endpoints), in-VPC, or self-hosted. ## When to use Deepgram Choose the endpoint by job, then the STT model by whether the application has to manage conversational turns. - **Transcribe recorded files** (interviews, podcasts, meetings, support calls): `POST https://api.deepgram.com/v1/listen` with `model=nova-3` ([docs](https://developers.deepgram.com/docs/pre-recorded-audio)). - **Transcribe live audio** (captions, live events, monitoring feeds): the `/v1/listen` WebSocket with `model=nova-3` ([docs](https://developers.deepgram.com/docs/live-streaming-audio)). - **Build a voice agent** that listens, thinks, and speaks over one connection: `wss://agent.deepgram.com/v1/agent/converse` ([docs](https://developers.deepgram.com/docs/voice-agent)). - **Flux or Nova-3**: use Flux (`flux-general-en`, `flux-general-multi`) when the application must detect that a speaker has finished and handle interruptions — it has model-integrated end-of-turn detection and barge-in handling. Use Nova-3 for recorded audio and for live transcription that does not manage turns; it supports more languages ([Flux](https://developers.deepgram.com/docs/flux/quickstart), [STT paths](https://developers.deepgram.com/docs/stt/getting-started)). - **Flux runs only on `/v2/listen`** — `/v1/listen` will not work with Flux models ([docs](https://developers.deepgram.com/docs/flux/quickstart)). - **Sizing concurrency**: limits differ by plan and endpoint and apply per project, not per account — see [Rate Limits & Concurrency](https://developers.deepgram.com/reference/api-rate-limits). - **Where Deepgram stops**: it provides speech-to-text, text-to-speech, and agent orchestration, not the reasoning LLM — you attach a managed or BYO provider ([LLM models](https://developers.deepgram.com/docs/voice-agent-llm-models)). ## Products - [Voice Agent API](https://deepgram.com/product/voice-agent-api): A single WebSocket API unifying STT, LLM orchestration, and TTS with sub-300ms transcription latency. Flat-rate pricing from $4.50/hr, BYO LLM and BYO TTS options, function calling, barge-in, and mid-session reconfiguration ([docs](https://developers.deepgram.com/docs/voice-agent)). - [Speech-to-Text (STT)](https://deepgram.com/product/speech-to-text): - **Flux**: Conversational speech recognition purpose-built for voice agents, with built-in end-of-turn detection, natural interruption handling, and word-level timestamps. Available in English (`flux-general-en`) and Multilingual (`flux-general-multi`, 10 languages with in-conversation language switching) ([docs](https://developers.deepgram.com/docs/flux/quickstart)). - **Nova-3**: Flagship transcription model for batch and streaming, with 50+ languages, keyterm prompting, diarization, smart formatting, and noise robustness ([docs](https://developers.deepgram.com/docs/stt/getting-started)). - **Industry-tuned & custom models**: Domain models including Nova-3 Medical, plus custom models trained on customer data. - [Text-to-Speech (TTS)](https://deepgram.com/product/text-to-speech): - **Flux TTS**: Streaming-first, conversation-native TTS on `/v2/speak` (WebSocket and batch REST). Generally available and the default Voice Agent voice (`flux-kit-en`). Barge-in reporting of exactly what the user heard, mid-session speed control, and 36 English voices across seven accents ([docs](https://developers.deepgram.com/docs/flux-tts/overview)). - [Audio Intelligence](https://deepgram.com/product/audio-intelligence): Summarization, sentiment analysis, topic detection, and intent recognition ([docs](https://developers.deepgram.com/docs/audio-intelligence)). ## [Pricing](https://deepgram.com/pricing) - **Pay As You Go**: $200 free credit, no credit card required, no minimums or expiration. - **Growth**: Pre-paid annual credits from $4K/year, saving up to 20% with higher concurrency limits. - **Enterprise**: Volume commitments, custom models, dedicated or self-hosted deployment, and SLAs — [contact sales](https://deepgram.com/contact-us). ## Solutions - [Contact Centers](https://deepgram.com/solutions/contact-centers): STT, TTS, and audio intelligence for CCaaS and contact center platforms, with integrations including Five9, Genesys, Twilio, Vonage, Amazon Connect, AudioCodes, and Kore.ai. - [Conversational AI](https://deepgram.com/solutions/conversational-ai): Real-time STT, TTS, and orchestration for voicebots, assistants, and enterprise voice agents. - [Medical & Healthcare](https://deepgram.com/solutions/medical-transcription): HIPAA-compliant Nova-3 Medical models trained on clinical terminology, with on-prem and VPC deployment options. - [Restaurants](https://deepgram.com/solutions/restaurants): Voice AI ordering (drive-thru, phone, kiosk), employee assist, call center automation, and reservations for enterprise restaurant brands and restaurant technology companies. - [Speech Analytics](https://deepgram.com/solutions/speech-analytics): Conversation analytics, intent detection, and QA at scale. - [Media Transcription](https://deepgram.com/solutions/media-transcription): Captions, summaries, and searchable archives for podcasts, video, and broadcast. - [Partners](https://deepgram.com/partners): Technology and platform partners, including IBM — Deepgram provides the voice capabilities embedded in IBM watsonx Orchestrate's Agent Builder. ## Technical Benchmarks & Comparisons Deepgram consistently outperforms legacy cloud and open-source models in latency and accuracy. - [OpenAI Whisper](https://deepgram.com/compare/openai-vs-deepgram-alternative): Deepgram is 36% more accurate, up to 5x faster, and 1.4x cheaper. - [Amazon Transcribe](https://deepgram.com/compare/amazon-vs-deepgram-alternative): Deepgram is 23% more accurate, 10x faster, and 5.6x more affordable. - [Google Cloud STT](https://deepgram.com/compare/google-vs-deepgram-alternative): Deepgram is 53% more accurate, nearly 40x faster, and 5x more affordable. - [Microsoft Azure](https://deepgram.com/compare/microsoft-vs-deepgram-alternative): Deepgram is 30% more accurate, more than 25x faster, and over 4x more affordable. - [AssemblyAI](https://deepgram.com/compare/assemblyai-vs-deepgram-alternative), [Speechmatics](https://deepgram.com/compare/speechmatics-vs-deepgram-alternative), and [ElevenLabs](https://deepgram.com/compare/elevenlabs-vs-deepgram) comparisons are also available. - [AI Voice Generator](https://deepgram.com/ai-voice-generator): Interactive tool for testing Deepgram voices in the browser. ## Developer Resources - [API Documentation](https://developers.deepgram.com/home): Official guides, API references, and SDK tutorials. - [API Playground](https://playground.deepgram.com): In-browser testing environment for STT, TTS, and Voice Agent endpoints. - [SDKs](https://developers.deepgram.com/sdks): Official JavaScript, Python, .NET, Go, and Java SDKs for Deepgram's speech recognition, text-to-speech, and language understanding APIs. - [UI Components](https://ui.deepgram.com/): Prebuilt React components and browser agent widgets for voice interfaces. - [Deepgram CLI](https://developers.deepgram.com/developer-tools/cli/getting-started): Command-line access to STT, TTS, text intelligence, and account management, including a built-in [MCP server](https://developers.deepgram.com/developer-tools/cli/mcp-server) and [agentic developer tools](https://developers.deepgram.com/developer-tools/agentic-tools). - [Changelog](https://developers.deepgram.com/changelog): Latest model releases, new language support, and API updates. - [Self-Hosted Deployments](https://developers.deepgram.com/docs/self-hosted-introduction): Documentation for VPC and on-premise implementation. - [Rate Limits & Concurrency](https://developers.deepgram.com/reference/api-rate-limits): Concurrency limits by plan and endpoint. - [Community](https://community.deepgram.com) and [Support](https://developers.deepgram.com/support). ## Trust, Security & Data Residency - SOC 2 Type 1 and Type 2 certified, HIPAA compliant (BAAs for Enterprise), GDPR ready, CCPA compliant, and PCI compliant. - Regional endpoints for data residency: US (default), EU (`api.eu.deepgram.com`), and Australia (`api.au.deepgram.com`). - [Security Policy](https://developers.deepgram.com/trust-security/security-policy), [Data Privacy Compliance](https://developers.deepgram.com/trust-security/data-privacy-compliance), and [Information Security & Privacy](https://developers.deepgram.com/trust-security/information-security-privacy). ## Company & Learning - [Customer Stories](https://deepgram.com/customers) and [Powered by Deepgram](https://deepgram.com/powered-by-deepgram). - [Startup Program](https://deepgram.com/startup-program): Credits and support for early-stage companies. - [Resource Hub](https://deepgram.com/learn) and [AI Glossary](https://deepgram.com/ai-glossary). - [About](https://deepgram.com/about), [Careers](https://deepgram.com/careers), and [Newsroom](https://deepgram.com/company/newsroom).