Google has released Gemini 3.5 Transcribe, a speech-to-text model designed for intelligent voice interactions that converts raw audio into accurate, polished text while handling background noise and disfluencies.
xAI has released Grok Voice Transcribe 2.0, a speech-to-text model that is twice as accurate as version 1.0 while maintaining the same pricing. Built on the audio foundation model behind Grok Voice, it leverages a unique dataset of live, noisy, multilingual audio to handle challenging real-world conditions like flaky phone lines and competing voices.
OpenAI has launched GPT-Live-1 in the API, providing developers with a single model capable of simultaneous listening and speaking to build voice-enabled applications. This release aims to simplify voice layer development by replacing traditional chained architectures with a unified approach that handles interruptions and context more naturally.
Google reports its technologies now support over 300 languages for more than 7 billion people, introducing new tools to improve speech understanding, data collection for underrepresented languages, and accessibility.
Google has introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, native speech-to-speech models designed for real-time voice agents that reason and execute tools without breaking conversational flow.
Google AI has released Gemini 3.5 Transcribe, a speech-to-text model available via two distinct API endpoints: the Interactions API for pre-recorded files and the Live API for real-time streaming.
NemotronLabs has introduced VoiceChat, an open-weight full-duplex speech-to-speech model designed to integrate listening, transcription, reasoning, and speaking within a unified streaming architecture. The system combines a streaming speech encoder and decoder-only language model with parallel specialized output streams for agent text and structured function calls, alongside an auxiliary RNN-T branch for incremental user transcription.
NemotronLabs has introduced NemotronLabs VoiceChat, an open-weight full-duplex speech-to-speech model that integrates native tool-calling capabilities within a unified streaming architecture. The system combines a streaming speech encoder and decoder-only language model with parallel output streams for agent text and structured function calls, alongside an auxiliary RNN-T branch for incremental transcription and a streaming TTS decoder.
SpaceXAI has released Grok Voice Transcribe 2.0, a speech-to-text model available via hosted API that claims twice the accuracy of version 1.0 at the same price point. The model targets difficult audio conditions such as noisy phone lines and competing voices, running in both batch and real-time streaming modes.
Google has released two new speech-to-speech models, Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, which are similar in shape to OpenAI's GPT-Live models.
An AI developer is asking for architectural advice to build a real-time voice AI system that achieves approximately 500 ms or less first-token latency while remaining cost-effective and scalable.
Aiden describes an architecture for agents that requires full-duplex voice conversation alongside complex visual reasoning and device actions, avoiding the limitation of a single model handling both. The solution splits responsibilities: a real-time voice model manages the conversation while a separate, stronger model executes background tasks, coordinating asynchronously via a task queue.
A user has released F-0310, a zero-dependency wrapper that enables fully local deployment of Coqui XTTS-v2 and Qwen 2.5 on Windows without requiring cloud APIs.
A new zero-dependency deployment wrapper named F-0310 enables fully local, offline inference for Coqui XTTS-v2 and Qwen 2.5 on Windows without requiring cloud APIs.
Paris-based voice AI company Gradium has launched Voice Design, a feature that generates new synthetic voices from written descriptions without requiring reference audio. The tool is now live in the Gradium API and Studio across five languages.
Researchers present TontaubeV1, a text-to-speech model that preserves natural prosody while enabling streaming inference from a single consumer GPU. The system uses a hierarchical DualCodec representation at 12.5 Hz to separate semantic streams from acoustic refinements, allowing for low-latency generation.
Researchers present TontaubeV1, a text-to-speech model that preserves natural prosody while enabling streaming inference from a single consumer GPU. The system uses a hierarchical DualCodec representation at 12.5 Hz to separate semantic streams from acoustic refinements, allowing causal decoding despite the noncausal nature of the underlying codec.
The AuK technical report introduces an open-source foundational model that unifies speech generation and editing through natural-language instructions and audio context.
Tencent has made its AuK 1.5B foundation model for speech generation and editing available as open-source software, including code and model weights.
Tencent has made the AuK foundation model for speech generation and editing available as open source, including code and weights on Hugging Face and ModelScope. The release includes AuK-Flash, a distilled variant designed for fast inference with only 4 steps.