Treble raises $18M for voice simulation platform

Granola, a meeting notetaker, is adding dictation capabilities to compete with Wispr, which recently added meeting notes to its dictation app. The move reflects intensifying competition in the voice AI space, with both companies expanding into each other's core functionality. This tit-for-tat feature development signals a consolidating market where meeting transcription and note-taking are becoming table stakes rather than differentiators.

Deepgram has acquired Daily, the maker of Pipecat, to strengthen its voice AI infrastructure capabilities. The deal gives Deepgram access to Pipecat's real-time voice and video infrastructure technology. The acquisition positions Deepgram to offer more comprehensive voice AI solutions by combining its speech recognition and processing capabilities with Daily's real-time communication infrastructure.

Nvidia released a free, open-source speech-to-text model capable of transcribing up to 40 languages. The model works on recorded audio but not live streams, positioning it as a tool for meeting and medical transcription while leaving real-time transcription to competitors like ElevenLabs.

Block has developed a voice model called Dessa that produces human-like speech output. Unlike typical speech models that rely on large language models for knowledge retrieval, Dessa derives its knowledge from conversational data it has been trained on. The development represents progress in voice AI technology, though specific performance metrics and deployment details remain limited in available information.