
Voice AI had a big week 🔥 It's taking on more of live calls, from reasoning while it speaks to preserving tone in translation and flagging synthetic callers. Here’s what stood out: - The AI audio market is projected to hit $117B by 2033, led by generative voice platforms. - @ElevenLabs released Scribe v3 Live, which uses the conversation so far to tell a pause from the end of a turn. - @retellai launched Tell-1, an in-house model designed to reason while speaking. Retell says it handles over 100M calls a month and has reached a $100M run rate. - @usebland introduced Tofu-1, a model built to listen and speak at the same time on banking and healthcare calls. - @DeepLcom introduced Aure, a translation model designed to preserve a speaker’s tone, emotion, and intonation across languages. - @ZyphraAI open-sourced ZONOS3-Mini, a speech model that can clone voices in 43 languages on-device. - @bandwidth previewed its own speech recognition model and deployed GPUs inside its phone network. - @modulate_ai added synthetic-caller detection to Velma alongside a new investment from Twilio Ventures. Full digest: voice-ai-newsletter.krisp.ai…










