What is Deepgram for?

Power enterprise voice solutions with Deepgram’s Speech-to-Text, Text-to-Speech, and Voice Agent APIs. Real-time, accurate, and built for scale.

There's a free tier to start with, and paid plans unlock the rest. Current pricing is on deepgram.com.

Deepgram News

Deepgram skills v1.7.0

October 12, 2026

Deepgram skills give AI coding assistants such as Claude Code, Cursor, and Codex up-to-date Deepgram API, SDK, and CLI guidance. The deepgram plugin now lists 14 skills, including the speech-to-text, text-to-speech, voice-agent, audio-intelligence, text-intelligence, browser-agent, and cli skills added in v1.6.0. v1.7.0 updates all 14 skills for the September API, SDK, and CLI changes, including Flux TTS pause markers and Early Access pronunciation overrides, the Flux STT Warning message and mid-stream numerals, the Voice Agent reusable-configuration and agent-variable endpoints, and FunctionCallCancelled. The skills also correct which SDKs support each feature and add the EU, Australia, and India regional endpoints. Install with npx skills add deepgram/skills or dg skills. For release details, see deepgram/skills v1.7.0.

Earlier releases: These releases, from September 14 through October 5

October 12, 2026

These releases, from September 14 through October 5, did not get their own changelog entries: Python SDK v7.10.0, JavaScript SDK v5.12.0, and Java SDK v0.10.1 add Voice Agent FunctionCallCancelled events and the defer_until_eot function setting. JavaScript streaming sockets also support for await...of. Python SDK v7.11.0, JavaScript SDK v5.13.0, and Java SDK v0.10.2 turn Numerals on or off during a Flux STT stream with Configure. Python SDK v7.12.0, JavaScript SDK v5.14.0, and Java SDK v0.11.0 add Flux TTS pause markers for batch synthesis and Early Access pronunciation overrides; see Speed, Pause, and Pronunciation. Python also adds typed Flux STT Warning messages. Go SDK v3.8.0 adds a Flux TTS client for batch and WebSocket synthesis, and ForceEndTurn for Flux STT. Go SDK v3.7.1 fixes Speak V1 WebSocket metadata and the Clear control message. .NET SDK v7.1.2 surfaces unknown WebSocket server frames through the Unhandled event and adds opt-in DeepgramWebSocketException errors for rejected WebSocket upgrades. Rust SDK v0.11.0 adds the rustls-tls-native-roots feature and Deepgram::tls_config for TLS-inspecting proxies and private CAs. This release includes a breaking change: a Flux

Python SDK v7.13.0

October 12, 2026

Deepgram Python SDK v7.13.0 adds connection.send_configure(ListenV1Configure(...)) to change keyterms and formatting options such as numerals on a live Listen V1 stream. Most Configure rejections arrive as typed ListenV1Error responses. Voice Agent sessions can exchange experimental custom Think-provider payloads with send_custom_to_think_provider(...) and AgentV1CustomFromThinkProvider responses, which require a configured custom Think-provider WebSocket endpoint. For release details, see deepgram-python-sdk v7.13.0.

JavaScript SDK v5.15.0

October 12, 2026

Deepgram JavaScript SDK v5.15.0 adds sendConfigure({ type: "Configure", ... }) for live Listen V1 streams, with most rejections delivered as typed ListenV1Error messages, and the experimental sendCustomToThinkProvider(...) for Voice Agent sessions. WebSocket connection errors now retry through the configured reconnectAttempts budget unless your application closed the socket. For custom transports, the retry budget resets after a connection stays open for the default five-second stable period. For release details, see deepgram-js-sdk v5.15.0.

Java SDK v0.11.1

October 12, 2026

Deepgram Java SDK v0.11.1 adds sendConfigure(ListenV1Configure) for live Listen V1 streams and onErrorMessage(...) for typed server errors. When no typed handler is registered, onError(...) receives a ListenV1ErrorException that carries the server error. Voice Agent sessions gain the experimental sendCustomToThinkProvider(...) and onCustomFromThinkProvider(...), which require a custom Think-provider wss:// endpoint. For release details, see deepgram-java-sdk v0.11.1.

Go SDK v3.9.0

October 12, 2026

Deepgram Go SDK v3.9.0 supports the current Voice Agent function-call protocol with typed function requests, responses, and cancellation events. Function settings accept arbitrary JSON Schema properties, and client-side functions can omit an endpoint. The Management client adds create, read, update, and delete endpoints for reusable Voice Agent configurations and Agent template variables. Configurations are stored and returned as JSON strings, and an update to a configuration replaces its metadata only. For release details, see deepgram-go-sdk v3.9.0.

.NET SDK v7.1.3

October 12, 2026

Deepgram .NET SDK v7.1.3 makes LiveSchema.Model optional on the Listen V1 WebSocket. Connect no longer throws NullReferenceException when you omit the model. Keyterms with an explicit non-Nova-3 model still fail validation with DeepgramException. For release details, see deepgram-dotnet-sdk 7.1.3.

Rust SDK v0.13.0

October 12, 2026

Deepgram Rust SDK v0.13.0 adds Text Intelligence analysis, SRT and WebVTT caption helpers, Voice Agent Flux and Aura speak settings with integer expressivity, typed Voice Agent server events, model listing that includes outdated models, streaming TTS options, named Aura-2 models, and self-hosted distribution credential management. Pre-recorded and streaming responses expose extra metadata, and Entity::raw_value preserves the spoken text before formatting. It also fixes a streaming worker that stopped when the idle timer fired without keep-alive, a Flux TTS deadlock on terminal errors under backpressure, and usage requests that omit the audio meter group. This pre-1.0 release includes breaking changes: Details::duration and Details::total_audio are now Option, and Details::channels and Details::streams are now Option. An absent value means the meter group was omitted, which is different from zero usage. Management and Auth::grant calls now use the configured base URL. To keep sending them to the hosted API from a custom-base-URL setup, create a separate client with Deepgram::new(api_key). A base URL with a path prefix must end in /. Streaming metadata and StreamResponse::TerminalRes

Rust SDK v0.12.0

October 12, 2026

Deepgram Rust SDK v0.12.0 fixes a hang on /v1/listen and /v2/listen streaming sessions when the response channel filled up, delivers the terminal transport error to your code, and preserves unknown streaming frames. WebSocket requests now send the full authority in the Host header. This pre-1.0 release includes a behavior change: sending data on an ended /v1/listen session now returns Err, where v0.11.0 returned Ok(()). For release details, see deepgram-rust-sdk 0.12.0.

Deepgram CLI v0.3.3

October 12, 2026

Deepgram CLI v0.3.3 adds two flags to dg speak. --play sends audio to your system player (ffplay, afplay, paplay, or aplay) and works with -o to save and play in one command. --list-voices lists the available text-to-speech voices and their languages. dg skills now installs Deepgram skills v1.7.0 from a pinned, hash-checked bundle, installs and removes only the skill folders the CLI owns, and cleans up skill files it can prove an earlier 0.3.x release wrote. For release details, see deepgram/cli v0.3.3.

Correction: over-limit Configure keyterm updates on /v1/listen

October 8, 2026

The October 2 entry on mid-stream Nova-3 keyterm updates said that a Configure message whose keyterms exceed the 500-token limit returns an Error and leaves the stream's keyterms unchanged. That is not the current behavior. An over-limit update stops transcription without an Error, and after about 30 seconds the server closes the stream with 1011 (NET-0000). Check the list size before sending the update. See Configure and Keyterm Prompting.

Nova-3 Medical: Improved English Models for Batch and Streaming

October 7, 2026

We have released improved Nova-3 Medical English models for batch and streaming, with fewer word errors and fewer errors on medical entities. Improved models Batch: nova-3-medical, English (en, en-US, en-AU, en-CA, en-GB, en-IE, en-IN, en-NZ) Streaming: nova-3-medical, English (en, en-US, en-AU, en-CA, en-GB, en-IE, en-IN, en-NZ) You do not need to change your API requests to access the improvements; the updates are live for all users. Available for hosted and self-hosted* customers. For more details, visit the Models & Languages Overview and Model pages. *For self-hosted customers, please request the updated models for download from your Deepgram account representative.

Nova-3 Medical Multilingual: Dutch, English, French, German, Hindi, Italian, Japanese, Portuguese, Russian, and Spanish

October 6, 2026

The nova-3-medical model now supports multilingual transcription with language=multi. Nova-3 Medical can now transcribe medical audio in which speakers switch between Dutch, English, French, German, Hindi, Italian, Japanese, Portuguese, Russian, and Spanish. Available for both batch and streaming. Availability Nova-3 Medical Multilingual is now available through our API. To access: Use model=nova-3-medical&language=multi in your API calls Available for hosted and self-hosted* customers Supports both pre-recorded (batch) and real-time streaming transcription Supported languages:Dutch (nl) English (en) French (fr) German (de) Hindi (hi) Italian (it) Japanese (ja) Portuguese (pt) Russian (ru) Spanish (es) For more details, visit the Models & Languages Overview and Model pages. *For self-hosted customers, please request the updated models for download from your Deepgram account representative.

Nova-3 Improved Models for Catalan, French, Indonesian, and Vietnamese

October 5, 2026

We have released improved Nova-3 monolingual models for four existing languages. These updates improve batch transcription for Vietnamese and Catalan and streaming transcription for French and Indonesian. Improved models Batch models Vietnamese (vi) Catalan (ca) Streaming models French (fr, fr-CA) Indonesian (id) You do not need to change your API requests to access the improvements to existing languages; the updates are live for all users. For more details and to see the full list of supported languages, visit the Models & Languages Overview page. For self-hosted customers, please request the updated models for download from your Deepgram account representative.

Update Nova-3 Keyterms Mid-Stream

October 2, 2026

Swap in the vocabulary for each step of a call without reconnecting. To replace a Nova-3 stream's keyterms during a /v1/listen session, send a keyterms array in a Configure message: { "type": "Configure", "keyterms": ["Deepgram", "customer service"] } Each keyterms array replaces the whole list, including keyterms set with the keyterm query parameter, and an empty array [] clears them. The 500-token keyterm limit that applies to the keyterm query parameter also applies to each update; an over-limit update returns an Error and leaves the stream's keyterms unchanged. The same message can also turn formatting features on or off, for example "features": { "numerals": true }. Mid-stream keyterms work on every Nova-3 streaming model, including monolingual languages, multilingual (language=multi), and Nova-3 Medical, and they behave the same as keyterms set when you open the stream. Updating Nova-3 keyterms with Configure on /v1/listen is available on the global endpoint (api.deepgram.com). It isn't available yet on the EU (api.eu.deepgram.com), Australia (api.au.deepgram.com), or India (api.in.deepgram.com) regional endpoints. Sending keyterms on another model returns an Error with code

Deepgram Self-Hosted October 2026 Release (261001)

October 1, 2026

Container Images (release 261001) quay.io/deepgram/self-hosted-api:release-261001 Equivalent image to:quay.io/deepgram/self-hosted-api:1.192.51 quay.io/deepgram/self-hosted-engine:release-261001 Equivalent image to: quay.io/deepgram/self-hosted-engine:3.135.0 Minimum required NVIDIA driver version: >=580, open kernel module flavor. See Drivers and Containerization Platforms for installation and verification steps. quay.io/deepgram/self-hosted-license-proxy:release-261001 Equivalent image to:quay.io/deepgram/self-hosted-license-proxy:1.11.1-1 quay.io/deepgram/self-hosted-billing:release-261001 Equivalent image to:quay.io/deepgram/self-hosted-billing:1.14.0-3 FIPS 140-3 Images FIPS 140-3 images are available for all four self-hosted services, published under the -fips tag suffix: quay.io/deepgram/self-hosted-api:release-261001-fips Equivalent image to:quay.io/deepgram/self-hosted-api:1.192.51-fips quay.io/deepgram/self-hosted-engine:release-261001-fips Equivalent image to:quay.io/deepgram/self-hosted-engine:3.135.0-fips quay.io/deepgram/self-hosted-license-proxy:release-261001-fips Equivalent image to:quay.io/deepgram/self-hosted-license-proxy:1.11.1-1-fips quay.io/deepgram/self-host

Flux TTS Inline Controls: Pause and Pronunciation

September 30, 2026

Flux TTS (/v2/speak) now supports inline pause and pronunciation controls in the text you send, alongside speed. Pause control. Insert silence with an escaped marker such as \{pause:1s\}: 500–3000 ms in 100 ms steps, up to 8 per request. Available on batch (REST) requests. Pronunciation control (Early Access). Override how a word is said with inline IPA, \{"word": "dupilumab", "pronounce": "duːˈpɪljuːmæb"\}, on both batch and streaming. Results can vary between generations while the feature is in Early Access, so generate each term several times before relying on it in production. Combination guardrails. Combinations that degrade audio are rejected with a clear error instead of producing bad audio: pronunciation cannot be combined with pause or with a speed other than 1.0, and speed is capped at 1.15 when a pause is present. Rejections return CONTROL_COMBINATION_INVALID or PAUSE_SPEED_CAP_EXCEEDED on batch, and ConfigureFailure or DATA-0002 on streaming. Reporting. Applied controls are counted in controls_applied on streaming and in the dg-pronunciations-applied and dg-breaks-applied headers on batch. See Speed, Pause, Pronunciation for syntax, limits, and error codes.

Toggle Numerals Mid-Stream on Flux STT

September 25, 2026

Switch Flux STT to digits for a PIN, phone number, or order number, then back to words, without reconnecting. To turn Numerals on or off during a Flux STT stream, send numerals as a boolean in a Configure message: { "type": "Configure", "numerals": true } The update applies to transcripts Flux STT sends after it processes the message, and ConfigureSuccess now includes numerals in the full active configuration it echoes. The numerals query parameter still sets the initial value when the stream opens. This replaces the connection-time-only behavior in the July 17 entry. Mid-stream numerals work on flux-general-en and on flux-general-multi for English, Spanish, French, German, Russian, Portuguese, Italian, and Dutch.

Nova-3 Improved Models for Flemish, German (Switzerland), Lithuanian, and Portuguese

September 24, 2026

We have released improved Nova-3 monolingual models for four existing languages. These updates enhance transcription quality for batch and streaming workloads. Improved models Batch and streaming models German (Switzerland) (de-CH) Portuguese (pt, pt-BR, pt-PT) Lithuanian (lt) Streaming models Flemish (nl-BE) You do not need to change your API requests to access the improvements to existing languages; the updates are live for all users. For more details and to see the full list of supported languages, visit the Models & Languages Overview page. For self-hosted customers, please request the updated models for download from your Deepgram account representative.

Nova-3 Improved Models for Danish, Estonian, Flemish, Italian, Lithuanian, Macedonian, Polish, Urdu, and Vietnamese

September 22, 2026

We have released improved Nova-3 monolingual models for nine existing languages. These updates enhance transcription quality for batch and streaming workloads. Improved models Batch and streaming models Polish (pl) Danish (da) Batch models Vietnamese (vi) Flemish (nl-BE) Streaming models Estonian (et) Macedonian (mk) Urdu (ur) Italian (it) Lithuanian (lt) You do not need to change your API requests to access the improvements to existing languages; the updates are live for all users. For more details and to see the full list of supported languages, visit the Models & Languages Overview page. For self-hosted customers, please request the updated models for download from your Deepgram account representative.

Correction: Browser Agent SDK has no client-side VAD

September 18, 2026

The Browser Agent SDK never shipped client-side Silero voice activity detection. Documentation published with the SDK described it as an available feature, and that documentation was wrong. The following did not exist in any released package: The vad option on MicrophoneOptions in @deepgram/agents. The speechThreshold and silenceThreshold VAD options. The speech-start and speech-end events on AgentMicrophone. The vad field in WidgetConfig and the widget's VAD configuration section. The @ricky0123/vad-web and onnxruntime-web peer dependencies. Uninstall them if you added them for VAD. MicrophoneOptions accepts sampleRate, echoCancellation, noiseSuppression, and autoGainControl. The microphone streams continuously while it is unmuted. Barge-in works, and it is driven server-side. The Voice Agent API emits user-started-speaking, which you pair with player.interrupt() to stop playback: session.on("user-started-speaking", () => player.interrupt()); @deepgram/react subscribes to that event and interrupts playback for you, so applications built on the provider and hooks need no change. The Browser Agent pages now reflect the shipped API. See JavaScript SDK, Widget Embedding Guide, and Rea

Correction: token minting uses ttl_seconds

September 18, 2026

The server-side token-minting example on the Browser Agent SDK overview posted {"ttl": 30} to POST /v1/auth/grant. The field is named ttl_seconds: body: JSON.stringify({ ttl_seconds: 30 }), /v1/auth/grant accepts unrecognized fields, returns 200, and issues a token with the 30-second default, so a request built from the old example raised no error while ignoring the requested lifetime. If you copied that example and increased the number, your tokens expired after 30 seconds. A 30-second token is enough for a full call. The token only has to be valid at the WebSocket handshake, and the connection stays open afterward. Minting a token requires an API key with Member permissions or higher. See Token-Based Authentication.

Nova-3 Pharma: Speech-to-Text for Pharmaceutical Use Cases (English)

September 17, 2026

We have released nova-3-pharma, a new Nova-3 model purpose-built for pharmaceutical vocabulary, with a focus on accurate drug-name recognition. It is designed for pharmacy and healthcare voice-agent workflows where getting the medication right matters most. Available in English for both batch and streaming. Availability Nova-3 Pharma is now available through our API. To access: Use model=nova-3-pharma in your API calls Available for hosted customersFor self-hosted customers, please request the model from your Deepgram account representative Supports both pre-recorded (batch) and real-time streaming transcription Supported languages: English (en, en-US, en-AU, en-CA, en-GB, en-IE, en-IN, en-NZ) For more details, visit the Models & Languages Overview and Model pages.

Deepgram Self-Hosted September 2026 Release (260915)

September 15, 2026

Container Images (release 260915) quay.io/deepgram/self-hosted-api:release-260915 Equivalent image to:quay.io/deepgram/self-hosted-api:1.192.37 quay.io/deepgram/self-hosted-engine:release-260915 Equivalent image to: quay.io/deepgram/self-hosted-engine:3.131.2 Minimum required NVIDIA driver version: >=580, open kernel module flavor The minimum driver version has increased this release, up from >=570.172.08. The Engine image is built against CUDA 13, and NVIDIA's official CUDA 13 support begins with the 580 driver branch. NVIDIA's documentation states the CUDA 13 minimum as >=580 rather than a specific version. The driver must be the open kernel module flavor, for example nvidia-driver-580-open. The proprietary build of the same version is not supported.On Blackwell hardware the consequence is total: NVIDIA never added Blackwell support to the proprietary 580 branch, so on that build a Blackwell GPU disappears from the system, nvidia-smi reports no devices, and Engine will not start. See Drivers and Containerization Platforms for installation and verification steps. quay.io/deepgram/self-hosted-license-proxy:release-260915 Equivalent image to:quay.io/deepgram/self-hosted-license-prox

India Endpoint Now Generally Available

September 15, 2026

The Deepgram India endpoint (api.in.deepgram.com) is now generally available for customers requiring data processing within India. Supported APIs The India endpoint supports the following Deepgram APIs: Speech-to-Text: /v1/listen and /v2/listen Text-to-Speech: /v1/speak Voice Agent: /v1/agent/converse Text Intelligence: /v1/read Configuration To use the India endpoint, replace api.deepgram.com with api.in.deepgram.com in your SDK or API requests. Your existing API keys and tokens work with the India endpoint. For WebSocket connections, use the corresponding India URLs: Speech-to-Text: wss://api.in.deepgram.com/v1/listen Text-to-Speech: wss://api.in.deepgram.com/v1/speak Voice Agent: wss://api.in.deepgram.com/v1/agent/converse For detailed configuration instructions and SDK examples, see our Regional Endpoints documentation. For what Deepgram guarantees about where your data is processed and stored, see Your Data at Deepgram.

SDK releases: Deepgram Python SDK v7.9.0, JavaScript SDK v5.11.0

September 14, 2026

Deepgram Python SDK v7.9.0, JavaScript SDK v5.11.0, and Java SDK v0.10.0 are now available. All three releases add Voice Agent ForceEndTurn support, Flux TTS expressivity controls for Agent providers, and the expanded Flux TTS speed range of 0.5 through 1.5 in 0.05 increments. Python SDK v7.9.0 Deepgram Python SDK v7.9.0 adds send_force_end_turn() and typed ForceEndTurn messages for Voice Agent sessions using a Deepgram V2 (Flux STT) listen provider. It also adds integer Agent Flux TTS expressivity from -2 through 2 and the expanded Flux TTS speed range. The unsupported aura-2-perseo-it model is removed from Speak V1 model literals; the API never served that model. For release details, see deepgram-python-sdk v7.9.0. JavaScript SDK v5.11.0 Deepgram JavaScript SDK v5.11.0 adds socket.sendForceEndTurn() and Agent Flux TTS expressivity from -2 through 2. The former SpeakV2Speed constants and Aura2PerseoIt model constants remain as deprecated aliases for source compatibility. For release details, see deepgram-js-sdk v5.11.0. Java SDK v0.10.0 Deepgram Java SDK v0.10.0 adds sendForceEndTurn(...), Agent Flux TTS expressivity, and the expanded Flux TTS speed range. This pre-1.0 release inc

@deepgram/react 0.2.0: @deepgram/react 0.2.0 adds runtime Voice Agent controls and makes React session lifecycle…

September 10, 2026

@deepgram/react 0.2.0 adds runtime Voice Agent controls and makes React session lifecycle handling safer. The release updates the React layer for @deepgram/agents 0.1.2 and @deepgram/sdk 5.9.0. Runtime controls and notifications Use sendAgentMessage() to inject an agent message with "default", "queue", or "interrupt" behavior. Use updateListen(), updateThink(), updateSpeak(), and updatePrompt() to change supported agent settings during a session. The provider and standalone hook now expose typed callbacks for protocol notifications, including confirmations for each runtime settings update. useAgentMode() and useDeepgramAgent() also expose the "thinking" mode and isThinking state. Lifecycle and upgrade notes Start, stop, reconnect, and React StrictMode handling now clean up canceled or failed microphone, playback, and client-tool resources. Automatic start failures call onSdkError, or console.error when no callback is set. A manual start() call rejects instead. This is a 0.2.0 release. Update exhaustive AgentMode switches to handle "thinking"; public context and hook result types add required members; registerClientTool() returns an unsubscribe function; and useDeepgramAgent().start

Hold a function call until the user's turn is confirmed

September 8, 2026

The agent begins building a reply before speech-to-text confirms the user has finished speaking. Agent audio is held until that confirmation, but function calls have always dispatched as soon as the LLM emitted them. For a function with an irreversible side effect, such as ending a call or booking an appointment, that meant the action could run for a turn the user then continued. defer_until_eot Add defer_until_eot: true to any function in agent.think.functions to hold it until the turn is confirmed. If the user keeps speaking and the turn resumes, the call is discarded before it does anything. { "name": "end_call", "description": "End the conversation and close the connection", "parameters": { "type": "object", "properties": {} }, "defer_until_eot": true } The default is false, so existing agents are unchanged. Deferring one function does not delay the others: a function that did not opt in still dispatches immediately, even with a deferred call ahead of it in the same turn. defer_until_eot applies to every listen provider, not only Flux. FunctionCallCancelled Clients now receive an explicit signal when a client-side function call they were sent is cancelled because the user start

Nova-3 Adds Kazakh, Plus Improved Models for Estonian, Hebrew, Latvian, Lithuanian, Macedonian, Malay, and Polish

September 3, 2026

We have added Kazakh as a new Nova-3 language and released improved Nova-3 monolingual models for seven existing languages. These updates expand language coverage and enhance transcription quality for batch and streaming workloads. New language Kazakh (kk, kk-KZ) — available for batch and streaming. Access this language by setting model="nova-3" and language="kk" in your request. Improved models Batch and streaming models Polish (pl) Hebrew (he) Batch models Macedonian (mk) Streaming models Estonian (et) Latvian (lv) Lithuanian (lt) Malay (ms) You do not need to change your API requests to access the improvements to existing languages; the updates are live for all users. For more details and to see the full list of supported languages, visit the Models & Languages Overview page. For self-hosted customers, please request the updated models for download from your Deepgram account representative.

Wider speed range for Flux TTS

August 31, 2026

Flux TTS speed now runs from 0.5 to 1.5 in 0.05 increments, up from 0.85–1.15. The default stays 1.0, and every previously accepted value still works. The wider range applies everywhere speed is accepted on Flux TTS: the /v2/speak streaming WebSocket (at connection and mid-session with Configure), the /v2/speak batch REST endpoint, and agent.speak.provider.speed in Voice Agent sessions running provider.version v2. A value outside the range is still rejected with SPEED_OUT_OF_RANGE, and one off the 0.05 increment with SPEED_INCREMENT_INVALID. Aura speed is unchanged at 0.7–1.5. For details, see Getting Started with Flux TTS, Client Messages, and TTS Models.

Flexible turn-taking control for Flux

August 28, 2026

Flux now gives you full control over how turns end. Beyond Flux's native end-of-turn detection, you can override it, suppress it entirely, or blend the two — and switch between these modes on the fly, even for a single turn, with a Configure message. Turn-taking goes from a fixed behavior to something you shape around each moment of a conversation. Three modes of turn-taking Automatic (default). Flux's native end-of-turn detection decides when a turn ends. EndOfTurn events carry "trigger": "model". Nothing changes for existing integrations. Semi-manual. Keep Flux's detection running, but override it whenever you have a definitive signal — a push-to-talk release, a DTMF tone, a "send" tap — by sending a ForceEndTurn. The model handles the ambiguous endings; you handle the unambiguous ones. Fully manual. Set eot_threshold=1.0 to suppress native detection entirely and own every turn boundary yourself, driving each ending with ForceEndTurn. Ideal when you already run your own VAD or endpointing stack. See Bring Your Own Turn Detection. Switch modes mid-stream — even per turn Because eot_threshold can be updated live with a Configure message, you can change turn-taking behavior for exac

Nova-3 Model Improvements for Bulgarian, Croatian, Estonian, Georgian, Italian, Latvian, Lithuanian, Malay, Marathi, and Telugu

August 28, 2026

We have released improved Nova-3 monolingual models for ten existing languages. These updates enhance transcription quality for batch and streaming workloads. Improved models Batch and streaming models Bulgarian (bg) Batch models Croatian (hr) Estonian (et) Italian (it) Latvian (lv) Lithuanian (lt) Malay (ms) Streaming models Georgian (ka, ka-GE) Marathi (mr) Telugu (te) You do not need to change your API requests to access these improvements; the updates are live for all users. For more details and to see the full list of supported languages, visit the Models & Languages Overview page. For self-hosted customers, please request the updated models for download from your Deepgram account representative.

Nova-3 Adds Assamese, Mongolian, and Pashto, Plus Improved Models for Czech, Danish, Swedish, Tagalog, and Turkish

August 27, 2026

We have added Assamese, Mongolian, and Pashto as new Nova-3 languages and released improved Nova-3 monolingual models for several existing languages. These updates expand language coverage and enhance transcription quality for batch and streaming workloads. 🆕 New languages Assamese (as, as-IN) — available for batch and streaming. Mongolian (mn) — available for batch and streaming. Pashto (ps, ps-AF) — available for batch and streaming. Access these languages by setting model="nova-3" and the relevant language code in your request. ✨ Improved models Batch and streaming models Czech (cs, cs-CZ) Danish (da, da-DK) Tagalog (tl) Batch models Turkish (tr, tr-TR) Streaming models Swedish (sv, sv-SE) You do not need to change your API requests to access the improvements to existing languages; the updates are live for all users. For more details and to see the full list of supported languages, visit the Models & Languages Overview page. For self-hosted customers, please request the updated models for download from your Deepgram account representative.

Deepgram Self-Hosted August 2026 Release (260826)

August 26, 2026

Container Images (release 260826) quay.io/deepgram/self-hosted-api:release-260826 Equivalent image to:quay.io/deepgram/self-hosted-api:1.192.31 quay.io/deepgram/self-hosted-engine:release-260826 Equivalent image to: quay.io/deepgram/self-hosted-engine:3.128.0 Minimum required NVIDIA driver version: >=570.172.08 quay.io/deepgram/self-hosted-license-proxy:release-260826 Equivalent image to:quay.io/deepgram/self-hosted-license-proxy:1.11.0-1 quay.io/deepgram/self-hosted-billing:release-260826 Equivalent image to:quay.io/deepgram/self-hosted-billing:1.14.0-1 FIPS 140-3 Images FIPS 140-3 images are available for all four self-hosted services, published under the -fips tag suffix: quay.io/deepgram/self-hosted-api:release-260826-fips Equivalent image to:quay.io/deepgram/self-hosted-api:1.192.31-fips quay.io/deepgram/self-hosted-engine:release-260826-fips Equivalent image to:quay.io/deepgram/self-hosted-engine:3.128.0-fips quay.io/deepgram/self-hosted-license-proxy:release-260826-fips Equivalent image to:quay.io/deepgram/self-hosted-license-proxy:1.11.0-1-fips quay.io/deepgram/self-hosted-billing:release-260826-fips Equivalent image to:quay.io/deepgram/self-hosted-billing:1.14.0-1-fips The

ForceEndTurn client message

August 26, 2026

You can now end the current user turn from your own signal instead of waiting for end-of-turn detection. ForceEndTurn Send ForceEndTurn to end the turn in progress immediately. The agent treats everything transcribed so far as the complete user utterance and moves on to think and speak. { "type": "ForceEndTurn" } Use it for push-to-talk release, DTMF tones, UI actions, or to drive turn taking from your own VAD or endpointing stack. ForceEndTurn requires a Deepgram V2 (Flux) listen provider (agent.listen.provider.version set to v2). With any other listen provider the server sends a FORCE_END_TURN_UNSUPPORTED warning and the turn does not end. For details, see Force End Turn. eot_threshold maximum raised to 1.0 agent.listen.provider.eot_threshold now accepts values up to 1.0. Set it to 1.0 to fully suppress natural end-of-turn detection and own turn taking with ForceEndTurn. For details, see Configure the Voice Agent and Bring Your Own Turn Detection.

Expressivity for Flux TTS voices, and two new OpenAI models

August 19, 2026

expressivity for Flux TTS agent.speak.provider.expressivity shifts a Flux TTS voice's delivery register along a calm to animated axis. It accepts the whole numbers -2 to 2 and defaults to 0, the voice's tuned delivery. Negative values produce calmer, steadier delivery; positive values produce more animated delivery with a wider pitch range. Every Flux voice supports it, and the value applies for the whole session. { "agent": { "speak": { "provider": { "type": "deepgram", "version": "v2", "model": "flux-haley-en", "expressivity": -1 } } } } expressivity is a beta parameter: its behavior may be tuned in future model versions, and non-default values raise the chance of hallucinations and pronunciation errors, so audition the value you plan to ship. 0 remains the only value validated for production. For value-by-value guidance, see Expressivity and Configure the Voice Agent. Two new OpenAI LLM models Deepgram's managed OpenAI provider adds two models: Model Pricing Tier gpt-5.6-luna Standard gpt-5.6-terra Advanced { "agent": { "think": { "provider": { "type": "open_ai", "model": "gpt-5.6-luna" } } } } For the full catalog, see LLM Models.

Deepgram CLI 0.3.0: Flux TTS and Flux STT

August 19, 2026

Deepgram CLI 0.3.0 is now available. It completes CLI support for Flux TTS, which reached general availability on 12 August, fixes Flux STT streaming, and enforces the CLI's exit-code contract. Upgrading from 0.2.26 picks up everything in this release. 0.2.27 was tagged but never published, so its changes arrive here. Full Flux TTS support on dg speak dg speak now defaults to flux-alexis-en on Speak v2 (WebSocket streaming) instead of Aura 2. Synthesised audio differs from earlier releases unless you request an Aura model explicitly: dg speak "Hello from Deepgram" -m aura-2-asteria-en -o hello.mp3 Two Flux-only controls are now available. Both are validated before the request and rejected for non-flux-* models. Option Values Notes --speed 0.85 to 1.15 in 0.05 steps 1.00 is nominal --expressivity -2 to 2 Beta. 0 is nominal; negative is flatter, positive more animated dg speak "So exciting!" --expressivity 2 --speed 1.05 -o lively.wav Any GA voice works with -m. For the full catalog see Voices & Languages; dg models lists Aura voices only. Flux streaming returns raw audio: linear16 by default, plus mulaw and alaw. The --container option and the wider --encoding set apply to Aura only

Nova-3 Adds Afrikaans and Georgian, Plus Improved Models for Hungarian, Macedonian, Russian, Slovak, Slovenian, and Urdu

August 17, 2026

We have added Afrikaans and Georgian as new Nova-3 languages and released improved Nova-3 monolingual models for several existing languages. These updates expand language coverage and enhance transcription quality for batch and streaming workloads. 🆕 New languages Afrikaans (af, af-ZA) — available for batch and streaming. Georgian (ka, ka-GE) — available for batch and streaming. Access these languages by setting model="nova-3" and the relevant language code in your request. Improved models Batch and streaming models Russian (ru) Slovak (sk) Slovenian (sl) Urdu (ur) Streaming models Hungarian (hu) Macedonian (mk) You do not need to change your API requests to access the improvements to existing languages; the updates are live for all users. For more details and to see the full list of supported languages, visit the Models & Languages Overview page. For self-hosted customers, please request the updated models for download from your Deepgram account representative.

Numerals Support Now Available for 4 New Languages: Bulgarian, Chinese (Cantonese, Traditional), Malay, and Korean (Monolingual Models)

August 14, 2026

Supported languages and language codes: Bulgarian (bg) Chinese (Cantonese, Traditional) (zh-HK) Malay (ms) Korean (ko, ko-KR) You can now use Deepgram’s Numerals feature with monolingual models for Bulgarian, Chinese (Cantonese, Traditional), Malay, and Korean. Numerals converts spoken numbers into digits (for example, "three hundred" → "300") in your transcript, helping you create more accurate and easily processed results. How to use Numerals: To enable numerals, add the numerals=true parameter to your Deepgram API request. Learn more about using Numerals and see the full list of supported languages on the Numerals documentation page.

Deepgram Self-Hosted August 2026 Release (260812)

August 12, 2026

Container Images (release 260812) quay.io/deepgram/self-hosted-api:release-260812 Equivalent image to:quay.io/deepgram/self-hosted-api:1.192.25 quay.io/deepgram/self-hosted-engine:release-260812 Equivalent image to: quay.io/deepgram/self-hosted-engine:3.127.0 Minimum required NVIDIA driver version: >=570.172.08 quay.io/deepgram/self-hosted-license-proxy:release-260812 Equivalent image to:quay.io/deepgram/self-hosted-license-proxy:1.11.0 quay.io/deepgram/self-hosted-billing:release-260812 Equivalent image to:quay.io/deepgram/self-hosted-billing:1.14.0 FIPS 140-3 Images FIPS 140-3 images are available for all four self-hosted services, published under the -fips tag suffix: quay.io/deepgram/self-hosted-api:release-260812-fips quay.io/deepgram/self-hosted-engine:release-260812-fips quay.io/deepgram/self-hosted-license-proxy:release-260812-fips quay.io/deepgram/self-hosted-billing:release-260812-fips These images run OpenSSL's FIPS 140-3 validated cryptographic provider. Enable FIPS by setting [fips] mode = "enabled" in each service's configuration. FIPS mode should be enabled only on the -fips images; the standard images do not include the FIPS provider. See FIPS-Compliant Deployment f

Flux TTS is Generally Available

August 12, 2026

Flux TTS, Deepgram's conversation-native text-to-speech model, is now generally available on /v2/speak — streaming WebSocket and batch REST. New at GA: Barge-in with exact spoken-text reporting.Interrupt cancels the active turn on the live socket, and SpeechInterrupted reports text_spoken and text_remaining — an exact answer to "what did the user actually hear," so your LLM context stays correct without reconstructing it from playback timing. See Interruption Handling. Mid-session speed changes.Configure adjusts speed without reconnecting, answered by ConfigureSuccess / ConfigureFailure. speed on both transports. One of 0.85–1.15 in 0.05 increments. An expanded voice catalog. 36 English voices across American, British, Irish, Australian, Indian, Singaporean, and Filipino accents, led by 12 featured all-rounders. See Voices & Languages. Also new alongside GA, but still in beta: an expressivity parameter that shifts a voice's delivery register from -2 (calm) to 2 (animated) on both transports. The tuned default (0) remains the recommended, production-validated setting. See Expressivity. Coming soon as fast follows: inline pause and pronunciation controls in Speak text — the controls_

Flux TTS is generally available, and is now the default agent voice

August 12, 2026

Flux TTS is the default agent.speak provider Flux TTS is generally available in the Voice Agent API. A Settings message that omits agent.speak altogether now uses Flux TTS with the flux-kit-en voice, where it previously used aura-asteria-en. Set agent.speak.provider.version to v2 with a flux-{voice}-{language} model to select Flux TTS explicitly, or v1 with an aura-* model for Aura. { "agent": { "speak": { "provider": { "type": "deepgram", "version": "v2", "model": "flux-alexis-en" } } } } The GA voice catalog is 36 English voices across seven accents. See the Flux TTS voice catalog. Breaking: default sessions requesting compressed audio are now rejected Flux TTS streams raw audio, so it accepts only the linear16, mulaw and alaw output encodings, with no container and no bit rate. Because it is now the default provider, a session that omits agent.speak and requests a compressed encoding (mp3, opus, flac, aac) or a container such as wav receives INVALID_SETTINGS where it previously received Aura audio. Name an aura-* model in agent.speak to keep the previous behavior: { "agent": { "speak": { "provider": { "type": "deepgram", "version": "v1", "model": "aura-asteria-en" } } }, "audio"

Nova-3 Adds Armenian, Plus Improved Models for Tamil, Indonesian, and Belarusian

August 10, 2026

We have added Armenian as a new Nova-3 language and released improved Nova-3 monolingual models for several existing languages. These updates expand language coverage and enhance transcription quality for both batch and streaming workloads. 🆕 New language Armenian (hy) — available for batch and streaming. Access it by setting model="nova-3" and language="hy" in your request. ✨ Improved models Batch models Tamil (ta) Indonesian (id) Streaming models Tamil (ta) Belarusian (be) You do not need to change your API requests to access the improvements to existing languages; the updates are live for all users. For more details and to see the full list of supported languages, visit the Deepgram Supported Languages documentation.

Improved Models Released for Multiple Languages

August 7, 2026

We have released improved Nova-3 monolingual models for the following languages. These updates enhance transcription quality and accuracy for both batch and streaming workloads. 🆕 Improved Models Batch models Bulgarian (bg) Bengali (bn) Bosnian (bs) Belarusian (be) Czech (cs) Norwegian (no) Streaming models Bulgarian (bg) Bosnian (bs) Czech (cs) Norwegian (no) Latvian (lv) You do not need to change your API requests to access these improvements; the updates are live for all users. For more details and to see the full list of supported languages, visit the Deepgram Supported Languages documentation.

Nova-3 Model Update

August 5, 2026

🌏 Nova-3 monolingual models now support Punjabi and Nepali with the following language codes: Punjabi: pa, pa-IN Nepali: ne Batch and streaming models are both available for these languages. Access this model by setting model="nova-3" and the relevant language code in your request. Learn more about Nova-3 and supported languages on the Models and Language Overview page.