Making audio more usable for humans and machines.
San Francisco
- Tweets 779
- Following 312
- Followers 1,310
- Likes 872
Pinned Tweet
Multi-Speaker 2.0 is live–the world’s first low and high-resolution multi-speaker separator. One recording of several people in. A clean labeled track per person out, including the moments when two people talk at once.
8 kHz to 48 kHz, one model. Confidence scores every 20ms.
2
4
161
"You have to hit between 10 and 15 milliseconds."
@themoko and @LoganatDell on why Dialogue RT–speech isolation for broadcasters runs on a Dell Pro Max with GB10 — and what a workstation unlocks that the cloud doesn't.
2
140
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
151
Real conversation isn't tidy. People interrupt, agree over each other, laugh through the answer.
This is the part every other tool gives up on. Multi-Speaker 2.0 keeps both voices as separate tracks, tells you how confident it is, and runs on anything from 8 to 48 kHz.
1
98
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
119
Most audio in the world was never recorded on separate tracks.
Jessica Powell breaks down source separation on Dell's #ReshapingWorkflows — how we pull vocals, instruments, or a single speaker out of a finished mix.
Full episode: podcasts.apple.com/us/podcas…
2
86
At IBC last week we demoed Dialogue RT: hand someone a mic in a chaotic room and they hear themselves come back clean, in real time.
It separates dialogue from crowd, music and room noise off a live feed in ~11ms. Not a noise gate — a separation model.
Reactions from NAB 👇
1
76
Cue sheets decide whether music royalties reach the songwriters and publishers who earned them. Incomplete data sends that money into the PRO "black box." AudioShake's music identification is now built into @MusicReportsInc' Cuetrak®.
1
1
54
In finished content, music is buried. @AudioShakeAI separates it out , then identifies each cue from the clean signal — even noisy, archival, or unlabeled tracks. Cuetrak turns that into a cue sheet. Read in @digitalmusicnws: digitalmusicnews.com/2026/09…
36
Dialogue RT is live on @Dell hardware at IBC today.
16 live broadcast feeds. One Dell Pro Max with GB10. Dialogue isolated at 11 ms, input to output.
Come see it → Dell, Stand 7.B47, Hall 7. RAI Amsterdam, through Sept 14.
1
1
97
Read more about our work with @Dell and how we brought Dialogue RT to broadcasters: dell.com/en-us/blog/more-tha…
1
41
Upload a file → get a cue sheet back.
@AudioShake’s copyright compliance system is now in @MusicReportsInc’ Cuetrak®. We separate the music out of the mix first, then identify each cue — so buried and unlabeled tracks still get caught.
🔗 audioshake.ai/press-releases…
65
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
One week to IBC2026. Dialogue RT's European debut.
Live mixed feed in. Dialogue stem and background stem out. 11 ms end to end.
Isolation, not suppression — every process downstream gets the element it needs instead of the same compromise.
Booth 14.F46.
1
2
76
When two people talk at once, diarization can label the moment — but a label isn't a track. The voices stay mixed, so those regions get thrown out.
Train on what's left and you get conversations where nobody interrupts.
Multi-Speaker 2.0 keeps both voices as audio.
1
96
The number we care about most is downstream. Against the strongest open-source separator, transcribing our separated tracks produces 4.1× fewer transcription errors.
2
1
158
Full methodology, datasets, and scoring configuration: audioshake.ai/post/multi-spe…
26
Multi-Speaker 2.0 is live–the world’s first low and high-resolution multi-speaker separator. One recording of several people in. A clean labeled track per person out, including the moments when two people talk at once.
8 kHz to 48 kHz, one model. Confidence scores every 20ms.
2
4
161
vs. the leading open-source diarizer: 30% fewer labeling errors, lower error on all 15 public test sets we ran, 38% better on telephone audio.
More on the technical achievements: audioshake.ai/post/multi-spe…
Samples: huggingface.co/spaces/audios…
1
68
Uses for Multi-Speaker Separation include:
Structured data creation
Transcription and captioning
Editing in film, TV, podcast, and radio
Full write-up: audioshake.ai/post/multi-spe…
53