jessica powell

I have no idea how this happened but...so excited and proud of our team!!! Audio nerds unite!

Full story here: audioshake.ai/post/how-espn-…

This work started through the @DisneyAccelerator, where we met the ESPN team. In the name of hot dog eating contests, slippery stairs races, and cornhole championships, onwards! 7/7

Today we're sharing our work with ESPN. I wanted to share what we've been building together — and how it involves T-Rex races, competitive sign spinning, and....Phil Simms! 1/7

"I'm going to Disney World" is almost synonymous with winning the @SuperBowl. @ESPN wanted the original Phil Simms clip for their 2026 ad — but the music was baked in. We pulled his voice from the mix. Watch: youtube.com/watch?v=w1l3z-…

Dialogue RT runs at 11ms end-to-end — the first AI dialogue isolation model to meet live broadcast latency requirements. Built for environments where the audio arrives messy and the pipeline can't wait!

Suppression attenuates unwanted sound within a mixed signal. Isolation extracts speech as its own signal, removing everything else. The result is a cleaner stem and independent control over dialogue & the rest of the mix — crowd, PA, ambient sound, etc. No guessing. No bleed.

@AudioShakeAI has just launched Dialogue RT — our real-time dialogue isolation model for live broadcast. Audio separation has always been about quality. This is about quality AND speed. A thread on why that matters and what we built. 🧵

Don't miss this free talk at SF Music Week on Friday 🎶 @AudioShakeAI Founder + CEO @themoko will be on stage at #SFMusicWeek for "Innovation and the Technology Shaping the Future of Music" alongside incredible minds from @noisepop, @chartmetric, @splashmusicco, + @sonos 🎼

Full technical breakdown: audioshake.ai/post/targeted-…

SAM is a genuinely exciting step forward, especially for creative exploration and research. At the same time, the benchmarks reinforce what we see every day in production: when accuracy, determinism, and real-world deployment matter, targeted models still set the standard.

@Meta new SAM Audio model is awesome—and it's great to see this level of momentum around audio after years of it being treated as a second-class modality. That said, the benchmarks tell a more nuanced story than the headlines.