PulseAugur
EN
LIVE 06:10:35

Together AI integrates Deepgram voice models, launches fast Whisper STT

Together AI has launched new speech-to-text (STT) and text-to-speech (TTS) capabilities, integrating Deepgram's advanced voice models and its own high-performance Whisper V3 API. This move aims to streamline the development of real-time voice agents by providing a unified platform for transcription, LLM processing, and synthesis. The offerings emphasize speed, accuracy, and enterprise-grade features like zero data retention and large file handling, addressing key latency and quality issues in current voice AI applications. AI

IMPACT Streamlines voice AI development by unifying STT, LLM, and TTS, addressing critical latency and quality issues for real-time applications.

RANK_REASON Product launch by a significant AI infrastructure provider, expanding its core offerings.

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 3 sources. How we write summaries →

COVERAGE [3]

  1. arXiv cs.AI TIER_1 English(EN) · So Kuroki, Yotaro Kubo, Takuya Akiba, Yujin Tang ·

    KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI

    arXiv:2510.02327v2 Announce Type: replace-cross Abstract: Real-time speech-to-speech (S2S) models excel at generating natural, low-latency conversational responses but often lack deep knowledge and semantic understanding. Conversely, cascaded systems combining automatic speech re…

  2. Together AI blog TIER_1 English(EN) ·

    Deepgram speech-to-text and voice models now available natively on Together AI

    Production STT and TTS from Deepgram, available on Together AI Dedicated Model Inference for real-time voice agents.

  3. Together AI blog TIER_1 English(EN) ·

    Together AI Launches Speech-to-Text: High-Performance Whisper APIs