Skip to navigation

Improved Nova-3 Models for German and Hindi

We’ve released improved Nova-3 monolingual models for German and Hindi. These updates enhance transcription quality for batch and streaming workloads.

Speech-to-Text

Deepgram CLI 0.5.0: dg speak gets --out-file, dg init needs --yes outside a terminal

Changes how dg speak saves audio, how dg init confirms setup outside a terminal, and what dg listen -o prints.

Developer Tools

Deepgram CLI 0.4.1: npm install, dg mcp drops SSE

Adds an npm install path, drops the dg mcp SSE transport, and narrows what a project config file can set.

Developer Tools

Improved Nova-3 Models for Catalan and Japanese

We’ve released improved Nova-3 monolingual models for Catalan and Japanese. These updates enhance transcription quality for batch and streaming workloads.

Speech-to-Text

Improved Nova-3 Models for Arabic, Armenian, Bengali, Bulgarian, Danish, English (Australia, New Zealand), Greek, Gujarati, Hebrew, Nepali, Slovak, Tagalog, and Urdu

We’ve released improved Nova-3 monolingual models for 13 existing languages. These updates enhance transcription quality for batch and streaming workloads.

Speech-to-Text

November 10, 2026

Deepgram Self-Hosted November 2026 Release (261110)

Container Images (release 261110)

  • quay.io/deepgram/self-hosted-api:release-261110

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-api:1.192.58-1
  • quay.io/deepgram/self-hosted-engine:release-261110

    • Equivalent image to:

      • quay.io/deepgram/self-hosted-engine:3.142.0-1
    • 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-261110

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-license-proxy:1.11.2-1
  • quay.io/deepgram/self-hosted-billing:release-261110

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-billing:1.15.1-1
Self Hosted

Verify Callback Requests with Basic Auth or Extra Metadata

The dg-token header is no longer a supported method to verify that callback requests come from Deepgram. If your callback server checks for dg-token, switch to Basic Auth. For Speech-to-Text, you can also use a secret passed with Extra Metadata.

See STT Callback, TTS Callback, and Text Intelligence Callback for details.

Speech-to-TextText-to-SpeechIntelligence

Nova-3 Multilingual Expands to 55 Languages and Adds Language Hinting

model=nova-3&language=multi now serves a new Nova-3 multilingual model that transcribes 55 languages with a single model. Code-switching and per-word language tags work as before, on the same /v1/listen endpoints, for both pre-recorded and streaming audio.

Key Features:

  • 55 languages, one model — See the Models & Languages Overview for the full list of languages and codes.
  • Code-switching — Transcribes speech that switches languages mid-conversation. When the model assigns languages, it returns them in languages on each alternative and in language on each word; on audio that switches languages, either field can be absent from a result.
  • Language hinting — The new optional language_hint parameter biases detection toward the languages you expect. See Nova-3 Language Prompting.
  • languages_hinted in responses — Every Nova-3 multilingual response now includes languages_hinted, the normalized hints you sent (en-GB is returned as en), or an empty array when you send no hint.
  • Same model string and endpoints — No changes to your requests are required.
Speech-to-Text

Japanese Numerals on Multilingual Models

The numerals parameter now formats Japanese (ja) on model=nova-3&language=multi, model=nova-3-medical&language=multi, and model=flux-general-multi. Japanese formatting already worked on monolingual requests. Hindi is not supported. See Numerals for the full list of 20 supported languages.

Speech-to-Text

Flux TTS Adds 41 Languages and 230 Voices

Flux TTS (/v2/speak) now speaks 41 more languages with 230 new voices, on the same streaming and batch endpoints with the same flux-{voice}-{language} model strings.

  • New languages. Bengali (bn), Bulgarian (bg), Catalan (ca), Chinese (zh), Croatian (hr), Czech (cs), Danish (da), Dutch (nl), Estonian (et), Finnish (fi), French (fr), German (de), Greek (el), Gujarati (gu), Hindi (hi), Hungarian (hu), Indonesian (id), Italian (it), Japanese (ja), Kannada (kn), Korean (ko), Latvian (lv), Lithuanian (lt), Malay (ms), Marathi (mr), Norwegian (no), Polish (pl), Portuguese (pt), Romanian (ro), Russian (ru), Slovak (sk), Slovenian (sl), Spanish (es), Swedish (sv), Tagalog (tl), Tamil (ta), Telugu (te), Turkish (tr), Ukrainian (uk), Urdu (ur), Vietnamese (vi). Browse every voice in the voice catalog. The English voices are unchanged. German, Dutch, French, Hindi, Italian, Japanese, Portuguese, Russian and Spanish ship with text normalization; the other 32 languages are in Early Access without it — numbers, dates and abbreviations are read as written, so write them out in words.
  • How to use. Pick a voice from the catalog and pass it as model. Streaming: wss://api.deepgram.com/v2/speak?model=flux-desislava-bg. Batch: POST https://api.deepgram.com/v2/speak?model=flux-desislava-bg.
  • Regions. The new voices are available on api.deepgram.com, api.eu.deepgram.com and api.in.deepgram.com. They are not yet available on the Australia endpoint (api.au.deepgram.com).
Text-to-Speech

Deepgram CLI 0.4.0: Browser Login Stores an Expiring Credential

Browser login now stores a time-limited credential, a breaking change for anyone who signed in before this release.

Developer Tools

October 20, 2026

Deepgram Self-Hosted October 2026 Release (261020)

Container Images (release 261020)

  • quay.io/deepgram/self-hosted-api:release-261020

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-api:1.192.55
  • quay.io/deepgram/self-hosted-engine:release-261020

    • Equivalent image to:

      • quay.io/deepgram/self-hosted-engine:3.140.1
    • 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-261020

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-license-proxy:1.11.2
  • quay.io/deepgram/self-hosted-billing:release-261020

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-billing:1.15.1
Self Hosted

Confirm Nova-3 Configure Updates

Know when a mid-stream update takes effect. Deepgram now answers every Configure message on a /v1/listen stream with exactly one ConfigureSuccess or ConfigureFailure message. When the update replaces the stream’s keyterms, ConfigureSuccess lists the keyterms now in effect:

{ "type": "ConfigureSuccess", "request_id": "550e8400-e29b-41d4-a716-446655440000", "keyterms": ["Deepgram", "customer service"] }
Speech-to-Text

Update Nova-3 Keyterms Mid-Stream on Every Regional Endpoint

Mid-stream keyterm updates with Configure on /v1/listen now work on the EU (api.eu.deepgram.com), Australia (api.au.deepgram.com), and India (api.in.deepgram.com) regional endpoints, as well as the global endpoint (api.deepgram.com). They behave the same on every endpoint.

Speech-to-Text

Update Nova-3 Keyterms Mid-Stream from the SDKs, Play Audio from the CLI, and Deepgram Skills v1.7.0

Deepgram Python SDK v7.13.0, JavaScript SDK v5.15.0, and Java SDK v0.11.1 let you reconfigure an active Nova-3 stream without reconnecting, including keyterms and numerals, and add custom Think-provider messages for Voice Agent sessions. Go SDK v3.9.0, .NET SDK v7.1.3, Rust SDK v0.12.0 and v0.13.0, Deepgram CLI v0.3.3, and Deepgram skills v1.7.0 are also available.

Developer Tools

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

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.

Speech-to-Text

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

We have released improved Nova-3 Medical English models for batch and streaming, with fewer word errors and fewer errors on medical entities.

Speech-to-Text

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

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.

Speech-to-Text

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

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.

Speech-to-Text