vincentmolina

Speaker Intelligence AI

[vincentmolina @VincentMolina86]

Life is about tipping points.
Today marks my first official day as CEO of pyannote AI. 🚀

So much has happened in just a few weeks.

[pyannoteAI @pyannoteAI]

🎟️ 2 free tickets to VapiCon 2026, Nov 11-12, SF. Reply with the speaker problem breaking your voice pipeline. We pick 2 winners on Oct 26. #VoiceAI #VapiCon2026

[Jyoti Bisht @joeyousss]

Lately @pyannoteAI ‘s diarization with MCP architecture has become my favourite topic to talk about !

[Riddhi @notYetConverged]

Haven’t worked with voice AI before, but following the @pyannoteAI Precision-3 release was surprisingly easy!

The diarization improvements & tunable speech & crosstalk sensitivity were interesting to learn about.

The upcoming Precision-4 features look exciting too. Can’t wait!

[pyannoteAI @pyannoteAI]

Speaker diarization (pyannoteAI), Transcription (Gladia), Compute (Modal): three layers of one Voice AI stack.
Oct 14, Paris. Panel + live demo of how they fit together.
https://t.co/h9XVYwfno8

[pyannoteAI @pyannoteAI]

Already running pyannote in production? On October 8 at 5 PM CEST, we'll show what changes when you switch to Precision-3: the new input parameters, per-frame probabilities in the output, and the migration steps.

Join live or get the recording ➡️ https://t.co/1NZzczCn5q

[pyannoteAI @pyannoteAI]

Last call: Paris Voice AI Meetup, Oct 14.
pyannoteAI x Gladia x Modal on running diarization, transcription, and infrastructure in production. Panel, Q&A, drinks.
https://t.co/h9XVYwfno8

[Hervé "pyannote" Bredin @hbredin]

Only two days after Precision 3 official release, first customers feedback is coming in.

💬 "1.8x speed improvement which puts it faster than our stt"
🗨️ "speaker over-splitting has gone down"
🗯️ "turn attribution is noticeably better"

So proud of the @pyannoteAI team!

[Baseten @baseten]

Many of our customers run workloads with speaker attribution, with the strictest demands on both quality and speed.

We partnered with the team at @pyannoteAI to deliver both: some of the highest-quality diarization models on the market, with 3.2x higher throughput and 9.6x lower latency.

[pyannoteAI @pyannoteAI]

Precision-3 is live. Diarization error down 10.4% vs Precision-2 (14.35 vs 16.02 DER across 15 benchmark datasets), with most of that gain coming from better speaker attribution: the model gets the wrong speaker 18.7% less often.

[vincentmolina @VincentMolina86]

@joeyousss @pyannoteAI Soooooooo thrillledddd to have you onboard !!! 🔥🔥🔥🤩🤩🤩

[Jyoti Bisht @joeyousss]

New in Voice AI : I am heading to lead Developer Relations at @pyannoteAI 🎉

Pyannote builds the speaker intelligence layer for voice AI: their own diarization models that get millions of downloads ( 5M last month)

Behind that number is a community: researchers, engineers, people building meeting tools and dubbing pipelines etc. Taking care of them is my job now 💚

A big big thank you to @paw_lean for sharing about this opportunity - and rooting for me! You’re a true community person in and out 💜

If you're building with voice, come say hi 👋🏻

[vincentmolina @VincentMolina86]

@iAligator 🤩

[Ali Haghani @iAligator]

3 big Circleback updates this month:

• a free plan with unlimited meetings
• import your past meetings from other apps
• an API for all your Circleback data

here's what's new from our team retreat on a ranch

[vincentmolina @VincentMolina86]

@garrytan Circleback is a great product indeed @iAligator 🤩

[Garry Tan @garrytan]

Circleback is so much better than Granola, it's not even close

Granola still doesn't even support multiple person disambiguation

[vincentmolina @VincentMolina86]

@baseten Exciting times ! @pyannoteAI @hbredin

[Hervé "pyannote" Bredin @hbredin]

@juliarturc @n0mad_0 @neilzegh Amazing work @juliarturc. We @pyannoteAI witness first hand that a lot of companies are currently training their own full duplex models as we provide them with extremely precise diarization of conversational training data. You should do a video about diarization by the way!

[vincentmolina @VincentMolina86]

@zachtratar @NotionHQ 🫡 @hbredin

[Zach Tratar @zachtratar]

Something new coming to @NotionHQ AI Meeting Notes very soon...

[Jerome Colombain | Monde Numérique @JeromeColombain]

La voix est en train de redevenir le canal le plus puissant de l’IA… et une startup française mène la révolution. 🔥
Vincent Molina, explique comment Pyannote reconnaît des voix jamais entendues à l’entraînement, pourquoi leur modèle surpasse les standards actuels, et comment il s’intègre dans des pipelines audio aux côtés de Whisper. 
👉 https://t.co/liaREi7SEp