Changelog

December 12, 2025

Aura-2 TTS Language Expansion

Deepgram has expanded Aura-2 (Text-to-Speech) to support the following languages:

  • Dutch
  • German
  • French
  • Italian
  • Japanese

Additionally, new voices have been added to the Spanish (es) model.

The expanded voice catalog spans genders, age groups, and speaking styles, supporting a wide range of enterprise use cases including customer service, healthcare, sales, interviews, and IVR.

You can explore all available voices, including featured voices, in the Voices & Languages section of our documentation and try them live in the Deepgram Playground.

December 11, 2025

PHI Redaction Now Available for Batch and Streaming Speech-to-Text

We’re excited to announce that PHI (Protected Health Information) redaction is now available for both batch (pre-recorded) and streaming speech-to-text.

redact=phi

You can now redact protected health information using the new phi parameter, which redacts the following entity types: condition, drug, injury, blood_type, medical_process, and statistics.

Key features:

  • Batch support: Available for all pre-recorded audio transcription
  • Streaming support: Available for real-time streaming transcription
  • Language support: Follows existing redaction language support (all languages for hosted batch, English only for streaming)
  • Combine with other redaction options: Use multiple redaction parameters together (e.g., redact=phi&redact=pci)

Example usage:

$curl \
>  --request POST \
>  --header 'Authorization: Token YOUR_DEEPGRAM_API_KEY' \
>  --header 'Content-Type: audio/wav' \
>  --data-binary @youraudio.wav \
>  --url 'https://api.deepgram.com/v1/listen?redact=phi'

For detailed information, see our Redaction documentation and supported entity types.

December 10, 2025

Container Images Release

Deepgram Self-Hosted December 2025 Release (251210)

Container Images (release 251210)

  • quay.io/deepgram/self-hosted-api:release-251210

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-api:1.172.2

  • quay.io/deepgram/self-hosted-engine:release-251210

    • Equivalent image to:

      • quay.io/deepgram/self-hosted-engine:3.104.10

    • Minimum required NVIDIA driver version: >=570.172.08

  • quay.io/deepgram/self-hosted-license-proxy:release-251210

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-license-proxy:1.9.2

  • quay.io/deepgram/self-hosted-billing:release-251210

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-billing:1.12.1

This Release Contains The Following Changes

  • Expands Nova-3 with 10 New Languages — Building on the 11-language expansion from the 251118 release, Nova-3 now supports 31 total languages. This release adds 10 additional languages, bringing improved accuracy and contextual understanding across:

    • Southern and Eastern Europe: Greek (el), Romanian (ro), Slovak (sk), Catalan (ca)
    • Northern and Baltic Europe: Lithuanian (lt), Latvian (lv), Estonian (et), Flemish (nl-BE), Swiss German (de-CH)
    • Southeast Asia: Malay (ms)

    Learn more in our announcement blogs: 10 new languages and previous 11-language expansion.

  • Adds Multilingual Keyterm Prompting for Nova-3 Multi — Nova-3 multilingual now supports multilingual keyterm prompting, allowing you to pass up to 500 tokens (~100 words) to boost recognition of brand names, industry jargon, proper nouns, and other mission-critical vocabulary across multilingual audio.

    This feature requires loading a newer version of the Nova-3 multilingual model. If you attempt to use keyterm prompting with an older version of the Nova-3 multilingual model, you will receive an error: Bad Request: The selected Nova-3 model does not support keyterm prompting. Contact Deepgram support for assistance with updating your model version.

    Learn more in the keyterm prompting documentation.

  • Improves Entity Formatting — Improves formatting for several entity types, including URLs and numeric entities that contain the word “thousand”.

  • Includes General Improvements — Keeps our software up-to-date.

December 3, 2025

EU Endpoint Now Generally Available

The Deepgram EU endpoint (api.eu.deepgram.com) is now generally available for customers requiring data processing within the European Union.

Supported APIs

The EU endpoint supports the following Deepgram APIs:

  • Speech-to-Text: /v1/listen and /v2/listen (excluding Whisper models)
  • Text-to-Speech: /v1/speak
  • Voice Agent: /v1/agent/converse
  • Text Intelligence: /v1/read

Configuration

To use the EU endpoint, simply replace api.deepgram.com with api.eu.deepgram.com in your SDK or API requests. Your existing API keys and tokens will work with the EU endpoint.

For detailed configuration instructions and SDK examples, see our Configuring Custom Endpoints documentation.

November 26, 2025

Nova-3 Multilingual Now Supports Keyterm Prompting

Keyterm Prompting has been expanded to include the Nova-3 multilingual model. Previously, this feature was only available for monolingual Nova-3 models — now you can use keyterms with both.

To enable it, simply use: model=nova-3&language=multi and include your keyterm list to boost recognition of domain-specific vocabulary such as brand names, proper nouns, and industry-specific terms.

For more details, see the Keyterm Prompting page.

November 24, 2025

Nova-3 Model Update

🎯 Nova-3 supports 10 new languages

We’ve added support for 10 new languages with non-English monolingual Nova-3 models. This continues our effort to significantly expand Nova-3 language support beyond English. The newly supported languages and their corresponding language codes are:

Newly Supported:

  • Catalan (ca)
  • Estonian (et)
  • Flemish (nl-BE)
  • German (Switzerland) (de-CH)
  • Greek (el)
  • Latvian (lv)
  • Lithuanian (lt)
  • Malay (ms)
  • Romanian (ro)
  • Slovak (sk)

Learn more about Nova-3 on the Models and Language Overview page.

November 18, 2025

Container Images Release

Deepgram Self-Hosted November 2025 Release (251118)

Container Images (release 251118)

  • quay.io/deepgram/self-hosted-api:release-251118

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-api:1.169.0

  • quay.io/deepgram/self-hosted-engine:release-251118

    • Equivalent image to:

      • quay.io/deepgram/self-hosted-engine:3.104.10

    • Minimum required NVIDIA driver version: >=570.172.08

  • quay.io/deepgram/self-hosted-license-proxy:release-251118

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-license-proxy:1.9.2

  • quay.io/deepgram/self-hosted-billing:release-251118

    • Equivalent image to:
      • quay.io/deepgram/self-hosted-billing:1.12.1

This Release Contains The Following Changes

  • Expands Nova-3 Monolingual Language Support — Nova-3 now supports 21 additional languages, bringing stronger accuracy and contextual understanding across:

    • Eastern Europe and Eurasia: Bulgarian (bg), Czech (cs), Greek (el), Hungarian (hu), Polish (pl), Romanian (ro), Russian (ru), Slovak (sk), Ukrainian (uk)
    • Nordics and Baltics: Estonian (et), Finnish (fi), Latvian (lv), Lithuanian (lt)
    • Western Europe: Catalan (ca), Flemish (nl-BE), Swiss German (de-CH)
    • South Asia: Hindi (hi)
    • East Asia: Japanese (ja), Korean (ko, ko-KR)
    • Southeast Asia: Malay (ms), Vietnamese (vi)

    Note: This release originally introduced 11 new languages (Bulgarian, Czech, Finnish, Hindi, Hungarian, Japanese, Korean, Polish, Russian, Ukrainian, and Vietnamese). The 10 additional languages from the 251210 release (Greek, Romanian, Slovak, Catalan, Lithuanian, Latvian, Estonian, Flemish, Swiss German, and Malay) are also supported via backward compatibility.

    Learn more in our announcement blogs: 11 new languages and 10 additional languages.

  • Adds 36-Language Detection Model — Adds support for a new language detection model that handles 36 languages. This feature requires enabling the use_v2_language_detection feature flag in the Engine TOML configuration. Language detection is available for pre-recorded audio only. Learn more in the language detection documentation.

  • Updates Status Endpoint — Updates the /v1/status endpoint to better reflect node startup and runtime state, preventing false Critical reports when the API starts before an Engine driver is ready. See the status endpoint documentation for the new status flow:

    • Initializing — Reported during node startup; transitions to Ready once initialization completes.
    • Ready — The node can service requests; transitions to Healthy after enough successful requests, or Critical if errors occur.
    • Healthy — Sustained success; can transition to Critical if failures arise.
    • Critical — Indicates node failures; can recover back to Ready once node can service requests again.

  • Enhances API Graceful Shutdown — Resolves an issue where the API container would not properly wait for outstanding work to complete before shutting down. The graceful shutdown period now defaults to approximately 10 minutes.

  • Improves Address Formatting — Improves formatting for street numbers in addresses.

  • Improves Aura-2 Latency Consistency — Improves latency consistency for Aura-2 text-to-speech requests.

  • Deprecates Legacy Intelligence Features — Legacy Intelligence features (analyze_sentiment=true, detect_topics=true, summarize=v1, and summarize=true v1 structure) are now deprecated in favor of newer versions. Requests using these parameters will return HTTP 400 errors. Migration guidance:

    • analyze_sentiment=true → use sentiment=true
    • detect_topics=true → use topics=true
    • summarize=v1 → use summarize=true or summarize=v2

    See the Speech-to-Text changelog for more details.

  • Includes General Improvements — Keeps our software up-to-date.

November 14, 2025

🚀 Introducing Saga Web: Voice-First AI Chat – Now Instantly in Your Browser

Transform How You Work: Real-Time Voice, Seamless Context, Full Command

We’re thrilled to announce Saga Web—the browser-based version of Saga, designed for instant, hands-free productivity. Whether you’re a longtime Saga power user or just getting started, Saga Web lets you:

  • Dictate and see your words appear with real-time speech-to-text. See Flux (Deepgram’s latest STT release!) in action – experience speech-to-text that feels instantaneous and magical.
  • Switch between voice and text on the fly. Never lose the thread or context whether you’re speaking or typing.
  • Directly execute actions and automate workflows from chat. Turn conversations into commands, without leaving your browser.
  • No downloads, no barriers. Access Saga from any device with just a click. Perfect for work, multitasking, and accessibility.

Saga Web is built to meet you where you are—unlocking voice-driven productivity for everyone, everywhere.

November 12, 2025

Use Deepgram’s Managed Cartesia TTS Models

We’re excited to announce an easier way to use Cartesia’s Text-to-Speech models inside Deepgram’s voice agent – Deepgram-managed Cartesia models.

Similar to our managed LLM’s, simply specify Cartesia as your model provider and the correct model name to get started instantly. No Cartesia account creation, setup, or payments are required – this feature is included as part of Deepgram’s Standard pricing tier.

For detailed information, please refer to our TTS documentation.

🤖 Claude Haiku 4.5 LLM Support

We’ve added support for Anthropic’s new Claude Haiku model in our Voice Agent API!

Implementation:

Configure your chosen model in your Voice Agent settings:

1{
2  "type": "Settings",
3  "agent": {
4    "think": {
5      "provider": {
6        "type": "anthropic",
7        "model": "claude-4-5-haiku-latest"
8      }
9    }
10  }
11}

For complete information about supported LLMs including Claude Haiku 4.5, visit our Voice Agent LLM Models documentation

November 10, 2025

Entity Detection Now Available for Streaming Speech-to-Text

We’re excited to announce that Entity Detection is now available for streaming (real-time) speech-to-text with Nova 3, Nova 2, Nova, and Enhanced models.

detect_entities=true

Previously available only for pre-recorded audio, you can now identify and extract over 50 unique entity types in real-time streaming transcriptions, including email addresses, names, locations, phone numbers, social security numbers, and more.

Key features for streaming Entity Detection:

  • Model support: Nova 3, Nova 2, Nova, and Enhanced models only (not available for Base models or Flux)
  • Real-time detection: Entities are included in final results (is_final: true messages)
  • Enhanced formatting: Includes both value (formatted) and raw_value (original spoken text) fields when formatting is enabled
  • Automatic inclusion: Empty entities array returned when no entities are detected

Example usage:

1const connection = deepgram.listen.live({
2  model: "nova-3",
3  language: "en-US",
4  smart_format: true,
5  detect_entities: true,
6});

For detailed information, see our Entity Detection documentation and supported entity types.