Inworld Realtime STT
Inworld Realtime STT is a realtime streaming STT API that understands users beyond their words. It combines low-latency speech recognition with voice profiling, extracting emotion, vocal style, accent, age, and pitch directly from raw audio so downstream LLMs and TTS systems can respond with more adaptive, expressive behavior. Developers can stream audio in real time, transcribe complete files, or extract voice profile signals through one unified API, with realtime bidirectional streaming over WebSocket, synchronous transcription for full audio files, voice profile signals on every streaming chunk, and multi-provider support through a single model ID. Every audio chunk can produce a realtime profile of the speaker with confidence scores, giving LLMs structured context such as whether a user sounds sad, frustrated, soft, high-pitched, or calm.
Learn more about Inworld Realtime STT
Grok Voice 3.0
Grok Voice 3.0 is a speech-to-speech model offered through the Grok Voice Agent API for developers building real-time voice applications. It accepts audio and text and returns spoken audio or text through a WebSocket session, supporting conversational turn-taking, interruptions, and function calling. Developers can connect the model to custom functions and server-side tools, including web search, X search, collections, and remote MCP servers. The API is intended for interactive voice agents such as customer-support and service-call experiences, and is also used for voice interactions in Tesla vehicles and Grok Bot. Billing is usage-based, with audio charged per minute and text input events billed separately. The API documentation includes setup instructions and examples for Python and JavaScript clients.
Learn more about Grok Voice 3.0
Phonic-1
Phonic-1 is a real-time speech-to-speech AI model offered through Phonic’s voice-agent development platform. It processes spoken input and generates spoken responses, supporting streaming conversations, configurable voices, turn-taking, and interruption handling. Developers can access the model through an API and SDKs, then use Phonic’s hosted tools to build, test, and deploy conversational voice agents. The platform is designed for developers and businesses creating voice applications for customer support, sales, appointment handling, and other spoken interactions. It includes options for hosted or self-managed deployment and provides documented integrations and webhook-based connections to external services. Phonic states that its voice technology supports more than 50 languages.
Learn more about Phonic-1
Dessa
Dessa is Block's speech-to-speech artificial intelligence model for conversational voice research and applications. It takes audio input and produces audio output directly, without converting speech into text as an intermediate step. Trained from scratch on human conversation rather than text, the model supports full-duplex interaction, listening and speaking at the same time, and runs locally on a MacBook Pro with an M4 Max chip or better. It captures conversational cues such as emotion, sarcasm, laughter, pauses, and turn-taking that can be lost in text-based pipelines. Dessa is intended for AI safety researchers and developers studying voice-model behavior, privacy, and real-time interaction. It is a speech model for research and development, not a general-purpose business application or hosted voice-agent platform.
Learn more about Dessa