Sonic splitter @AudioshakeAi. Author of The Big Disruption. Fiction in @nytimes @VICE @WIRED. Mostly noise, some signal.
San Francisco
- Tweets 337
- Following 1,019
- Followers 4,855
- Likes 10,234
Pinned Tweet
I have no idea how this happened but...so excited and proud of our team!!! Audio nerds unite!
Speechless (insert vocal isolation joke). Thank you @TIME for recognizing AudioShake’s stem separation technology as one of the Best Inventions of 2023. More on why in the year of AI everywhere, our work stood out: time.com/collection/best-inv…
6
23
3,978
Fun fact for the audio nerds: live broadcast gives you roughly 10–15 milliseconds to process audio before latency becomes a problem.
That’s one reason source separation has historically lived in post-production. Unlike noise reduction, you can actually separate the speech from the background, BUT it’s been too slow.
We’ve been working on changing that at @AudioShakeAI. Dialogue RT separates speech from background audio in 11 ms end to end.
So out of post and into live.
Kinda wild to watch individual sounds disappear in real time.
1
4
237
jessica powell retweeted
Dialogue RT isolates dialogue from a live feed in 11 ms end to end — measured model input to isolated output.
That's not noise suppression. It's two stems: dialogue, and everything else.
On NVIDIA DGX Spark and Blackwell architecture GPUs. audioshake.ai/post/isolating…
1
1
149
jessica powell retweeted
Four days at IBC, hundreds of conversations. The question we heard most:
“Can it run on the feed, not the file?”
Answer: yes. Live, less than 11 milliseconds. Another great year in the books.
1
1
117
jessica powell retweeted
Here's where to find us at IBC all week!
AudioShake — 14.F46 (live Dialogue RT demos)
AI-Media – 5.C34
Dell — 14.D43
Ortana — 1.C37h
ScorePlay — 1.D47
Telos Alliance — 8.D37
Book time now: tidycal.com/audioshake/ibc26
1
1
98
Today we're releasing @AudioShakeAI Multi-Speaker 2.0: one recording of several people talking goes in, a clean labeled track per person comes out — including the moments they talk over each other.
2
2
6
280
7/
Every job also returns confidence scores, per 20 ms frame and per file: how sure we are the speech went to the right speaker, and how cleanly the overlapping voices came apart.
Rank a million hours, skip the unusable parts.
1
30
8/
Try it free on-demand or via API:
audioshake.ai/products/multi…
27
Coming out of X hibernation to unveil our new brand vision for @AudioShakeAI , courtesy of my 12-yr-old.
P.S. We have a bet on whether I can get more likes than his Brawl Star YT videos, so help me out here.
1
5
212
jessica powell retweeted
"Music is more durable than more strictly cognitive tasks like coding."
Why? The core of music lives in live performance, learning an instrument, and the artist-fan connection. The stuff that's hardest to automate is exactly the stuff we value most.
Great clip from @trapital 👇
1
3
118
When the recording is the only copy that exists, you can't let a tool guess. Suppression smears the words. Generative enhancement invents them. So @AudioShakeAI built Speech Recovery to do neither — it isolates the speech that's actually there, and adds nothing.
1
2
529
More about @AudioShakeAI 's Speech Recovery Models: audioshake.ai/post/introduci…
1
2
72
There is so much media & archives blocked from streaming & social platforms bc of copyright compliance. Distribution demand has multiplied, and the compliance workflows underneath it have not.
Today we're launching the system that closes that gap. tinyurl.com/2nyduuy9
1
1
163