Deepgram Expands Nova-3 with Spanish, French, and Portuguese Support

Building on our global momentum, Nova-3 now brings its enterprise-grade accuracy, adaptability, and scalability to three of the world’s most widely spoken languages: Spanish, French, and Portuguese. Together, these languages represent billions of speakers across Europe, the Americas, Africa, and beyond, unlocking opportunities for enterprises with truly global ambitions.

In August, we announced Nova-3 support for German, Dutch, Swedish, and Danish, marking a pivotal leap forward in European coverage. Today, we are continuing that momentum by enabling organizations to deliver voice AI experiences in global bridge languages that connect continents, cultures, and industries.

Nova-3 now supports:

• Spanish: es, es-419
• French: fr, fr-CA
• Portuguese: pt, pt-BR, pt-PT

Each of these languages brings its own unique challenges for speech-to-text:

• Spanish: Wide dialect diversity across Spain, Latin America, and the Caribbean. A fast, syllable-timed rhythm makes accurate transcription harder than in stress-timed languages like English.
• French: Complex connected speech with liaison and elision, frequent homophones, and nasal vowels that make word boundaries difficult to detect.
• Portuguese: Significant differences between Brazilian and European Portuguese in phonetics and vocabulary. Nasalization and vowel reduction add complexity for ASR systems.

By extending Nova-3 to these dialects, Deepgram ensures enterprises can handle this linguistic diversity with accuracy, speed, and confidence.

While every Nova-3 release improves overall accuracy, these three languages are especially impactful because of their global scale. Enterprises can now deploy voice AI across:

• Spanish-speaking regions: Contact centers in Mexico and Spain, banking compliance in Latin America, healthcare documentation across the U.S. and Latin America.
• French-speaking regions: Canadian bilingual enterprises, government and legal transcription in France, customer support operations across Africa.
• Portuguese-speaking regions: Brazilian fintech and call centers, healthcare in Portugal, retail across Lusophone Africa.

Nova-3 ensures:

• Keyterm Prompting captures domain-specific terminology across diverse dialects without retraining.
• High accuracy in noisy, real-world environments such as call centers, drive-thrus, and clinical settings.
• Scalable real-time performance for mission-critical enterprise deployments.

Nova-3 delivers double-digit relative WER reductions compared to Nova-2 across all three languages, with streaming performance leading the way. Portuguese and Spanish in particular show standout gains, with streaming WER improvements exceeding 20 percent.

Key highlights:

• Portuguese streaming: 24.35 percent relative WER improvement (largest gain to date)
• Spanish streaming: 21.40 percent relative WER improvement
• French streaming: 14.26 percent relative WER improvement despite high linguistic complexity

These results underscore Nova-3’s ability to adapt to diverse phonetics and dialects while maintaining enterprise-grade accuracy.

Switching to Spanish, French, or Portuguese in Nova-3 is as simple as adding a language parameter to your request. For example, here is how to transcribe Spanish audio with Nova-3 using cURL:

curl --request POST \
  --header "Authorization: Token YOUR_DEEPGRAM_API_KEY" \
  --header "Content-Type: audio/wav" \
  --data-binary @youraudio.wav \
  "https://api.deepgram.com/v1/listen?model=nova-3&language=es"

Replace language=es with es-419, fr, fr-CA, pt, pt-BR, or pt-PT depending on your use case.

Explore more options in the Models & Languages Overview.

With Spanish, French, and Portuguese now part of Nova-3’s expanding portfolio, Deepgram strengthens its mission to make enterprise voice AI both globally accessible and locally precise. These three additions not only extend coverage to billions of speakers but also underscore our commitment to pushing the boundaries of speech accuracy across the world’s most demanding languages.

Sign up free and unlock $200 in credits - enough to power over 750 hours of transcription or 200 hours of speech-to-text in Spanish, French, or Portuguese. Explore more details on our Models & Language Overview page and dive into Nova-3’s world-class accuracy and adaptability across global markets.

Join a growing community of innovators: your feedback will directly influence upcoming language support, helping shape the next wave of Deepgram’s global voice AI.

Stay in the loop: Subscribe now to receive priority announcements on new features and language expansions, as Deepgram continues to scale globally with confidence.