←
Soniox577 posts
Soniox avatar✓

Soniox ✓

@soniox_ai

Low-latency real-time speech-to-text, text-to-speech and translation APIs.

◷ Joined March 2022▧ 159 Photos and videos
Soniox avatar
Soniox@soniox_ai·Nov 3

Introducing Soniox STT v6, the world’s first speech recognition model that knows what your voice agent is saying and automatically filters out the agent’s own voice. Two connected innovations make this possible: → Agent text input gives recognition context from both sides of the conversation, improving its understanding of user responses. → Agent voice echo cancellation filters out the agent’s playback while continuing to listen to the user, even when they interrupt. The result: fewer false interruptions, fewer clarification loops, and more reliable conversations. V6 also delivers significantly higher accuracy across 60+ languages for voice typing, meeting transcription, multilingual recognition, and real-time translation, with stronger speaker diarization, language identification, endpoint detection, and context customization. One unified speech model. All 60+ languages. The full feature set, with languages freely mixed. Available now for real-time streaming and asynchronous transcription. Explore what’s new in Soniox STT v6: soniox.com/blog/soniox-stt-v…

1:04
Soniox avatar
Soniox@soniox_ai·Nov 13

Last week, we launched Soniox STT v6. Customers are already running it in production and reporting real improvements across voice agents, voice typing, and AI meeting products. Pipecat’s open-source benchmark confirms it: v6 delivers the lowest semantic word error rate and ranks among the fastest models. Two world-first innovations in speech-to-text: → Agent text input gives v6 context from both sides of the conversation, improving recognition of short answers, names, and numbers. → Agent voice echo cancellation filters out the agent’s own voice, reducing false interruptions while continuing to recognize the user when they interrupt. Voice typing is faster and more accurate, with fewer corrections to numbers, dates, names, and specialized terms. Meeting transcription and speaker diarization are stronger, helping applications capture what was said and who said it through interruptions, overlapping speech, and language switching. Real-time speech-to-text translation also improves, preserving meaning and critical details such as dates, amounts, and identifiers. One unified speech model. Native-speaker accuracy across 60+ languages. Available for real-time streaming and asynchronous transcription. Explore Soniox STT v6: soniox.com/blog/soniox-stt-v…

0:41
Soniox avatar
Soniox@soniox_ai·Nov 12

Soniox STT does not stop at transcription. It also supports real-time speech-to-text translation, delivering the transcript and the translation together in one stream across 3,600 language pairs. Translate too early and you get the meaning wrong. Translate too late and the conversation moves on. With v6, Soniox STT knows when to start translating, so meaning stays accurate without falling behind. The details hold up too: dates, amounts, and currencies stay accurate, even when people mix languages in the same sentence. Build dedicated speech translation apps, or add live translation to your existing meeting product. Explore real-time translation: soniox.com/blog/soniox-stt-v…

0:40
Soniox avatar
Soniox@soniox_ai·Nov 10

Good AI meeting notes start with getting two things right: what was said and who said it. Soniox STT v6 improves both across 60+ languages. We trained v6 for real meetings: people interrupt, talk over each other, and switch languages mid-sentence. More accurate transcription and speaker diarization help your app capture the right decision and assign the action item to the right person. Give v6 your team’s vocabulary, including participant names, product names, and internal terms, to improve recognition further. Build live captions, speaker-labeled transcripts, and AI meeting notes with Soniox STT v6. Explore Soniox STT v6: soniox.com/blog/soniox-stt-v…

0:29
Soniox avatar
Soniox@soniox_ai·Nov 9

Soniox STT v6 is the most accurate real-time STT in the world right now, according to @pipecat_ai's benchmark. And because we are also fast, the top 3 models on the Pareto frontier are all Soniox real-time STT models. Soniox STT v6 hit a 0.71% pooled semantic word error rate and delivered final transcripts a median of 266 ms after the speaker stopped. Do not take any benchmark at face value. Test providers side by side on your own audio: soniox.com/compare-stt

Post attachment
Soniox avatar
Soniox@soniox_ai·Nov 6

Soniox STT v6 delivers our most accurate and fastest voice typing yet across 60+ languages. More accurate where it matters: numbers, currencies, units, dates, names, alphanumeric identifiers, and specialized terminology. For v6, we invested heavily in real-world dictation data across applications, professional vocabulary, speaking styles, and languages. The result is significantly higher accuracy for the speech your users actually dictate. Developers can further improve accuracy by supplying context such as product names, professional vocabulary, and project terminology. Fewer corrections. Less friction. Voice input people can rely on. Build voice typing into editors, messaging apps, forms, medical software, and enterprise tools across all 60+ languages. Explore voice typing with Soniox STT v6: soniox.com/blog/soniox-stt-v…

0:30
Soniox avatar
Soniox@soniox_ai·Nov 5

Your voice agent should never interrupt itself. Soniox STT v6 is the world’s first speech-to-text model with agent voice echo cancellation. When the agent’s playback reaches the microphone, STT can mistake it for user speech. The agent interrupts itself, responds to its own words, or enters a feedback loop. V6 uses agent text input capability to distinguish the agent’s playback from user speech and automatically filter it out of recognition. Crucially, it keeps listening while the agent speaks. When the user interrupts, v6 recognizes their speech while filtering out the agent’s voice. Fewer false interruptions. Fewer feedback loops. More reliable voice conversations. Across speakerphones, laptops, in-car systems, and other devices where playback reaches the microphone. Available across all 60+ languages. Explore agent voice echo cancellation: soniox.com/blog/soniox-stt-v…

0:32
Soniox avatar
Soniox@soniox_ai·Nov 4

Speech recognition has been missing half the conversation. Soniox STT v6 is the world’s first STT model that knows what your voice agent is saying. Introducing agent text input. Developers can stream the agent’s outgoing text into STT as it is sent to TTS, giving recognition context from both sides of the conversation. The question asked, the options offered, and the details being confirmed all help the model recognize the user’s response. V6 uses that context to recognize names, numbers, specialized terms, and short answers significantly more accurately. Fewer clarification loops. Less frustration. More reliable voice agents. Available across all 60+ languages. Explore agent text input: soniox.com/blog/soniox-stt-v…

0:30
Soniox avatar
Soniox@soniox_ai·Oct 29

Soniox STT and TTS share support for the same 60+ languages. STT can recognize language switches mid-sentence, and TTS can naturally mix languages in its responses. For developers building voice agents, this means consistent multilingual language coverage for both sides of the conversation, all through one provider.

0:15
Soniox avatar
Soniox@soniox_ai·Oct 27

Soniox speech AI focuses on native-speaker accuracy, including regional accents, local names, and everyday speech. Many STT models list hundreds of languages, but accuracy determines what you can reliably build and deploy. Build for the way your customers actually speak. soniox.com/platform

Soniox avatar
Soniox@soniox_ai·Oct 22

Soniox TTS v2 delivers natural, high-quality speech at a price built for scale. Power 24/7 voice agents that handle millions of customer interactions without compromising on quality or cost efficiency. Try it live: soniox.com/text-to-speech

0:14
Soniox avatar
Soniox@soniox_ai·Oct 21

Soniox TTS RT v2 is in the best-performing quadrant of @cekuraAi TTS Model Benchmarks, among 26 TTS models. Factor in price and what that means for scaling, and we believe Soniox TTS takes the top spot. 🚀 Explore the benchmark: benchmarks.cekura.ai/text-to…

Post attachment
Soniox avatar
Soniox@soniox_ai·Oct 14

Soniox real-time STT lets you tune endpoint detection to avoid cutting callers off mid-sentence without adding unnecessary latency. For voice agents, the right endpoint settings can help balance natural turn-taking with low latency. Test it on your own calls: soniox.com/speech-to-text

0:38