Weekly Digest: Updates from Meta, AssemblyAI, Apple, Deepgram, and others

Let’s watch this once more. The possibilities of <2 seconds translation are incredible.

Top Updates 💪

• Meta AI launches Audiobox, the first foundational audio model supporting both voice and text prompts

• AssemblyAI raised $50M Series C round to build more Speech AI models

• Twilio exits from Video SDK and moves it to Zoom

• Introduced in iOS 17, Apple's Personal Voice feature lets you record your voice and then use it to speak what you type during a phone call

• Deepdub, startup with ties to HBO Max and Fox, launches royalty program for AI voice clones

• Deepgram launches Aura, a fast conversational TTS for real-time AI agents

---

Noteworthy 📝

• Cloud AI and Local AI could coexist - per Windows’s Chief

• How Do Noise Canceling Headphones Work?

• Zoom fatigue is not burnout. It’s boreout. New study: when meetings are virtual, we’re not overwhelmed—we’re understimulated.

• Resemble AI launches new real-time Deepfake audio detector

• Read AI expands to scheduling, project management, audio recap

• Respeecher raises $1M to add a few studios to its media and gaming clients

• MacWhisper transcription now up to 3x faster on Macs with Apple silicon

• SoundHound AI acquires SYNQ3 to expand its Customer Service solutions and create the largest Voice AI provider for restaurants

• How to scale your voice assistant: Interview with FedEx’s Paul Pugal

• Whispp to launch new real-time assistive Voice technology, helping millions with voice disabilities speak in their natural voice

---

Demos 😎

• Dubbing AI: Real-Time Voice Changer app

• Meta AI translation demo

• Lip sync + voice cloning + subtitles

• Xound AI sound enhancement system

• Google AI presents Translatotron 3: A novel, unsupervised Speech-to-speech translation architecture

• How to convert PDFs into Audiobooks using OpenAI’s TTS API