Hervé "pyannote" Bredin @hbredin Twitter profile | Pikagi

Pikagi

Hervé "pyannote" Bredin

@hbredin

1,609

Followers

693

Following

106

Media

821

Statuses

Hervé Bredin /👨🏼‍🔬 Researcher @CNRS @IRIToulouse @SAMOVAIRIT / ⚒️ Co-founder and CSO @pyannoteAI / 👨🏻‍💻 Creator of 🎹 pyannote

https://t.co/bJznXfDmUZ

Joined March 2009

Don't wanna be here? Send us removal request.

Pinned Tweet

@hbredin

Hervé "pyannote" Bredin

20 days

Hola 🎹 #pyannote speaker diarization aficionados! @pyannoteAI is hiring a developer advocate 🥑 to * manage the open source (OSS) community * create content for OSS & @pyannoteAI users * provide support to OSS & @pyannoteAI users Ping me if interested.

1

2

9

Last Seen Profiles

@GHS_law

@airstreet

@snaksnok

@Joe_Nute

@bokeplokalmalam

@nunuchantique

@Petria_Thompson

@gekidanshibainu

@orca193

@BangashSak31571

@ZllSen

@UncutLoc

@Rumi_Toybox

@pennys2pounds

@GonnNSFW

@matsu_yukie

@VmaniakJ

@poetrystudios20

@Refai4A

@Gaa_Kaa

@Sindetrella

@queenly_owl

@BowBrunil

@sex58558849

@minjeongsgirl

@MouChenghao

@K2319400

@Khushhalkasi

@18dart5

@DreamWorksTV

@Lukasluan191

@AlexLindback

@yejisita0_o

@fabien_gay

@St_Lukes_SE25

@j_ashworth656

@hbredin

Hervé "pyannote" Bredin

8 months

#pyannote mood

Tweet media one

9

8

137

@hbredin

Hervé "pyannote" Bredin

2 years

🎹pyannote·audio 2.0 is now officially out!

Tweet media one

4

15

104

@hbredin

Hervé "pyannote" Bredin

2 years

🎹 pyannote + 🗒 notebook = pyannotebook pyannotebook is a custom @ProjectJupyter widget built on top of #pyannote .core and #wavesurferjs . It can be used to visualize and edit temporal audio labels, without leaving the notebook.

Tweet media one

3

13

85

@hbredin

Hervé "pyannote" Bredin

3 years

I discretely released a few 🎹pyannote․audio pretrained pipelines in the last few weeks. Try them out in the dedicated @huggingface @streamlit space:

Tweet media one

2

19

85

@hbredin

Hervé "pyannote" Bredin

10 months

Just released version 3.1 of #pyannote speaker diarization toolkit 🥇Same accuracy. 🏝️Less dependency hell. ⚡️Probably faster. Try it here and please RT 🙏

Tweet card media

Pretrained pipelines - a Hugging Face Space by pyannote

4

21

81

@hbredin

Hervé "pyannote" Bredin

2 years

#pyannote models are now #2 and #3 most downloaded audio models on @huggingface hub. Yet, I am struggling finding applicants for the two open positions to make it even better (2 years research engineer, 3 years PhD scholarshop) Can anyone help?

Tweet media one

@hbredin

Hervé "pyannote" Bredin

2 years

🫂Enjoying #pyannote speaker diarization pipelines? 👥What if it could do speaker separation as well? 🗣️Share the word if you want to make it happen! 🥳We are hiring! Twice! 👇

1

10

25

1

24

63

@hbredin

Hervé "pyannote" Bredin

1 year

Something's cooking 👩‍🍳 #pyannote

Tweet media one

2

3

44

@hbredin

Hervé "pyannote" Bredin

1 year

I am considering starting an open-source business around #pyannote open-source state-of-the-art speaker diarization toolkit. Please help me make the right decisions by either filling this form and/or retweeting: A few promising stats:

3

16

39

@hbredin

Hervé "pyannote" Bredin

4 years

In the spirit of reproducible (and reusable) research, the best model is now available in @PyTorch torch.hub and pyannote․audio, making speaker verification as easy as this few lines of code:

Tweet media one

2

12

37

@hbredin

Hervé "pyannote" Bredin

2 years

Now available in a model hub near you! Go get it! It's going to make some noise (pun intended) @huggingface #pyannote .audio #proudsuperviser

Tweet card media

pyannote/brouhaha · Hugging Face

@LavechinMarvin

Marvin Lavechin

@LavechinMarvin

2 years

If you ever need to detect audio segments that have been recorded in noisy or reverberant environments, Brouhaha might be worth checking! 🎙️🥁🚨🔊 Submission to #ICASSP2023 co-authored with Marianne Métais led by @hbredin 🧵⬇️

3

24

82

1

9

36

@hbredin

Hervé "pyannote" Bredin

2 years

The little things...

Tweet media one

1

1

35

@hbredin

Hervé "pyannote" Bredin

3 years

Pretrained model is available on @huggingface hub!

Tweet card media

pyannote/segmentation · Hugging Face

@ArxivSound

arXiv Sound

3 years

``End-to-end speaker segmentation for overlap-aware resegmentation. (arXiv:2104.04045v1 []),'' Hervé Bredin, Antoine Laurent,

0

2

12

2

6

36

@hbredin

Hervé "pyannote" Bredin

1 year

Took a bit more than a month but here it is: #pyannote 3.0.0. Pretrained pipeline should be much (MUCH) better than 2.1.1! #pyannote

Tweet card media

pyannote/speaker-diarization-3.0 · Hugging Face

@hbredin

Hervé "pyannote" Bredin

1 year

… coming soon in an open source library near you 🫣

1

0

1

1

11

34

@hbredin

Hervé "pyannote" Bredin

2 years

"How I reached 1st place at Ego4D 2022, 1st place at Albayzin 2022, and 6th place at VoxSRC 2022 speaker diarization challenges" 😎 Short answer: with 🎹 #pyannote .audio Long answer:

1

6

34

@hbredin

Hervé "pyannote" Bredin

1 year

#pyannote @jsalt2023

Tweet media one

1

4

33

@hbredin

Hervé "pyannote" Bredin

1 year

Irish #pyannote

Tweet media one

@ISCAInterspeech

INTERSPEECH 2024

@ISCAInterspeech

1 year

Really feeling "The Cocktail Party Effect" with all the buzz tonight at the Guinness Storehouse! #INTERSPEECH2023

Tweet media one

Tweet media two

Tweet media three

Tweet media four

1

0

27

2

0

33

@hbredin

Hervé "pyannote" Bredin

1 year

Announcing "Powerset multi-class cross entropy loss for neural speaker diarization", first paper by my PhD student @SAMOVAIRIT @IRIToulouse Alexis Plaquet to appear at #Interspeech2023 TL;DR: #pyannote gets better 👇

3

6

33

@hbredin

Hervé "pyannote" Bredin

1 year

Tweet media one

5

1

31

@hbredin

Hervé "pyannote" Bredin

1 year

#pyannote ranks 12th ... just behind "small" organisations such as @OpenAI , @GoogleAI , @facebook , @microsoft , @runwayml , @StabilityAI ... and with "slightly" less million dollars.

@osanseviero

Omar Sanseviero

1 year

Check out this @huggingface organizations leaderboard created by a community member!🤯 - @StabilityAI is the org with the most likes (15k) - @MetaAI is among the orgs with most models (700!) - @runwayml has the highest avg of downloads per model

Tweet media one

8

31

151

1

0

31

@hbredin

Hervé "pyannote" Bredin

1 year

Just because **I** put some open-source code on the internet, that **YOU** chose to use, does not make **ME** responsible for solving **YOUR** problems. There, I said it. #opensource

3

0

31

@hbredin

Hervé "pyannote" Bredin

2 years

🎹 "One speaker segmentation model to rule them all" ... in which you'll learn about one of the most popular audio 🎧 model available on @huggingface hub #pyannote .audio #speakerdiarization

One speaker segmentation model to rule them all

CNRS / IRIT / SAMoVA

herve.niderb.fr

1

7

28

@hbredin

Hervé "pyannote" Bredin

11 months

When did that happen? Why didn’t anyone tell me? #pyannote

@BenjaminDEKR

Benjamin De Kraker 🏴‍☠️

11 months

@jeffrschneider Diarization is a solved problem, I believe. (Just not currently being used in the ChatGPT app)

1

0

2

4

1

29

@hbredin

Hervé "pyannote" Bredin

6 years

I just released version 1.0 of library for neural speaker diarization. This includes a complete rewrite in @PyTorch , faster training, and a tunable speaker diarization pipeline. Check it out:

Tweet card media

GitHub - pyannote/pyannote-audio: Neural building blocks for speaker diarization: speech activity...

Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding - GitHub - pyannote/pyannote-audio: Neural build...

3

11

26

@hbredin

Hervé "pyannote" Bredin

2 years

Yet another @openai Whisper + #pyannote .audio hack. But this time, it's mine!

Tweet card media

rich-transcription.ipynb

rich-transcription.ipynb. GitHub Gist: instantly share code, notes, and snippets.

gist.github.com

3

5

26

@hbredin

Hervé "pyannote" Bredin

6 years

Automatically estimating the learning rate bounds for the "1cycle policy" cc @GuggerSylvain @jeremyphoward

Tweet card media

Estimating the learning rate bounds for the "1cycle policy"

Estimating the learning rate bounds for the "1cycle policy" - Estimating the learning rate bounds for the "1cycle policy".ipynb

gist.github.com

1

8

26

@hbredin

Hervé "pyannote" Bredin

1 year

Woot! Woot! #pyannote team ranked third… without ensembling 😇

Tweet media one

@huh_jaesung

JaesungHuh

1 year

The VoxSRC workshop will start within 10 minutes!!!!

Tweet media one

4

0

7

1

1

25

@hbredin

Hervé "pyannote" Bredin

2 years

🫂Enjoying #pyannote speaker diarization pipelines? 👥What if it could do speaker separation as well? 🗣️Share the word if you want to make it happen! 🥳We are hiring! Twice! 👇

1

10

25

@hbredin

Hervé "pyannote" Bredin

3 years

Took a while but pyannote․audio now provides a decent speaker embedding model.

Tweet media one

2

2

25

@hbredin

Hervé "pyannote" Bredin

1 year

Getting ready for #Interspeech2023 #pyannote #MyAcademicLife #marketting

Tweet media one

Tweet media two

1

0

23

@hbredin

Hervé "pyannote" Bredin

1 year

My talk about speaker diarization at @jsalt2023 summer school is now available online #pyannote #SpeakerDiarization #FrenchAccent

Tweet card media

Speech segmentation and speaker diarization

Lecture presented by Hervé Bredin (IRIT). This course introduces basis knowledge on speech segmentation. Processing a full recording, obtained for instance from a TV or radio show, requires to...

umotion.univ-lemans.fr

0

3

23

@hbredin

Hervé "pyannote" Bredin

2 years

Yet another @openai Whisper + #pyannote .audio hack. I’m seeing a trend here…

@PodscriptAI

PodScript - Podcast Transcripts

2 years

Coding has been ongoing for a few days. Great progress today 🥳 I guess I have a solution for speaker diarization and some of the timestamp mismatches when the speaker changes. #MachineLearning #coding #Python

Tweet media one

0

0

4

1

4

21

@hbredin

Hervé "pyannote" Bredin

5 years

Looks like I am one of the lucky ones!

@GoogleAI

Google AI

5 years

Congratulations to the recipients of the 2019 Google Faculty Research Awards! We received over 900 proposals from universities worldwide, covering diverse research areas such as #MachineLearning , #HumanComputerInteraction , #DistributedComputing and more.

11

138

657

3

0

23

@hbredin

Hervé "pyannote" Bredin

4 years

Current status: converting pyannote․audio to @PyTorchLightnin Smaller, more readable codebase and (hopefully) faster training!

Tweet media one

0

0

22

@hbredin

Hervé "pyannote" Bredin

1 year

Now that #pyannote 3.0.0 is out, grab your pretrained "powerset" speaker segmentation on @huggingface hub while it's hot!

pyannote/segmentation-3.0 · Hugging Face

@hbredin

Hervé "pyannote" Bredin

1 year

Announcing "Powerset multi-class cross entropy loss for neural speaker diarization", first paper by my PhD student @SAMOVAIRIT @IRIToulouse Alexis Plaquet to appear at #Interspeech2023 TL;DR: #pyannote gets better 👇

3

6

33

0

7

23

@hbredin

Hervé "pyannote" Bredin

1 year

6 months later…

Tweet media one

@hbredin

Hervé "pyannote" Bredin

2 years

The little things...

Tweet media one

1

1

35

3

0

21

@hbredin

Hervé "pyannote" Bredin

10 months

What about #pyannote ? Now do: * downloads = f(money raised) * downloads = f(head count) 🫣

Tweet media one

@EMostaque

Emad

10 months

Stability AI is the leader in open AI models. Open models trained on our cluster are the most downloaded globally (below @huggingface ) & with what we have coming up by us & ecosystem it’s going to be a fun year ahead Open ftw Super proud of team, onwards 🚀 🌟

Tweet media one

39

65

622

1

2

22

@hbredin

Hervé "pyannote" Bredin

2 years

Looking for 6️⃣ pyannote·audio (paid 💰) interns 👨🏻‍🎓 in Toulouse ☀️, France 🥖🍷🧀 📣 Spread the word!

Tweet card media

6 (paid) pyannote.audio internships in Toulouse, France · Issue #1092 · pyannote/pyannote-audio

More details here.

1

10

21

@hbredin

Hervé "pyannote" Bredin

1 year

#pyannote /segmentation model is downloaded more than 2M times every month on @huggingface model hub. Who are you? @julien_c @reach_vb is it possible to get access to a more detailed breakdown of this number?

Tweet media one

2

0

22

@hbredin

Hervé "pyannote" Bredin

1 year

The word is out: I do scientific consulting too.

0

2

20

@hbredin

Hervé "pyannote" Bredin

3 years

🎇 Upload an audio file and get voice activity and overlapped speech detection directly in the browser! Thanks, @huggingface Spaces!

@hbredin

Hervé "pyannote" Bredin

3 years

Pretrained model is available on @huggingface hub!

2

6

36

0

7

20

@hbredin

Hervé "pyannote" Bredin

1 year

On my way to #Interspeech2023 #MyAcademicLife #Killarney

Tweet media one

2

1

20

@hbredin

Hervé "pyannote" Bredin

10 months

#pyannote is in da place ⁦ @joinstationf ⁩

Tweet media one

0

3

21

@hbredin

Hervé "pyannote" Bredin

1 year

I’ll be in Dublin for #Interspeech2023 between August 20th and August 25th. Any #pyannote user willing to give back to the 'community' by buying #pyannote maintainers a beer, a dinner, or a proper hotel room in Dublin? DMs open. #AskingForAFriend #MyAcademicLife #OpenSource

6

2

20

@hbredin

Hervé "pyannote" Bredin

1 year

Oyez, oyez, (future) #pyannote users at #Interspeech2023 Talk to me at this afternoon first poster session, (literally) interact with the poster, and win a free #pyannote sticker in return!

Tweet media one

1

5

20

@hbredin

Hervé "pyannote" Bredin

2 years

Anyone around here interested in a (paid 💵) gig around 🗣️speaker diarization? You would port 🎹pyannote·audio pretrained pipelines on mobile phones. Please retweet if you'd like to see 👀 it happen!

Tweet card media

Pretrained pipelines - a Hugging Face Space by pyannote

2

8

18

@hbredin

Hervé "pyannote" Bredin

1 year

PSA: fixed a major #reproducibility issue with A100 @NVIDIA GPUs In case you tried #pyannote pretrained pipelines in the past on such devices and were disappointed by the accuracy, please give it another try with the latest commit in develop branch.

Tweet card media

Diarization pipeline output variability with different CUDA versions. · Issue #1370 · pyannote/py...

Hi, I just want to report I observed quite a large variability in the output of the pyannote diarization pipeline if I change the CUDA version on my side. For example, switching between torch 1.13....

1

3

18

@hbredin

Hervé "pyannote" Bredin

9 months

PSA: I just released #pyannote 3.1.1 that fixes a major issue with the option to control the number of speakers. If you did not use the option before, you were not impacted by this bug.

Tweet media one

2

2

19

@hbredin

Hervé "pyannote" Bredin

11 months

What if #pyannote could do separation too? 🔀 I am hiring a postdoc to help me with that! 👩‍🎓 Apply here: … or please RT 🙏 #diarization #pyannote

0

10

18

@hbredin

Hervé "pyannote" Bredin

2 years

Would anyone use this?

Tweet media one

@hbredin

Hervé "pyannote" Bredin

2 years

Any @ProjectJupyter and #wavesurferjs expert around here? I'd like to build a Jupyter interactive widget to visualize and edit (!) #pyannote .audio annotation instances directly in a notebook but have no idea where to start...

Tweet media one

3

1

15

3

2

15

@hbredin

Hervé "pyannote" Bredin

2 years

Yet another @OpenAI Whisper + #pyannote .audio hack!

@dwarkesh_sp

Dwarkesh Patel

2 years

I've been using OpenAI's Whisper model to generate initial drafts of transcripts for my podcast. But Whisper doesn't identify speakers. So I stitched it to a speaker recognition model. Code is below in case it's useful to you. Let me know how it can be made more accurate.

Tweet media one

42

66

846

1

1

17

@hbredin

Hervé "pyannote" Bredin

4 years

Started storing and versioning (public and private) pyannote․audio models on "The Hub"... and loving it!

@huggingface

Hugging Face

4 years

🤯The Hub is now open to all models, not just transformers! Open-source, open-sharing, open-science for the win. This redesign was led by @victormustar ➡️

4

26

91

1

3

17

@hbredin

Hervé "pyannote" Bredin

4 years

pyannote․audio + prodigy = ✨

@explosion_ai

Explosion 💥

4 years

🔊 The audio & video UI lets you label segments or transcribe. We even have experimental workflows for speaker diarization with a model in the loop, thanks to the awesome @hbredin and #pyannote .audio!

1

1

10

1

3

16

@hbredin

Hervé "pyannote" Bredin

2 years

Any @ProjectJupyter and #wavesurferjs expert around here? I'd like to build a Jupyter interactive widget to visualize and edit (!) #pyannote .audio annotation instances directly in a notebook but have no idea where to start...

Tweet media one

3

1

15

@hbredin

Hervé "pyannote" Bredin

1 year

Would you pay for #pyannote custom or premium models? #AskingForAFriend (I'd appreciate a few ReTweets 🙏)

Yes

18

No

28

Show me the results

53

6

27

15

@hbredin

Hervé "pyannote" Bredin

8 months

#pyannote in the wild! Let's go Desh!

Tweet media one

@rdesh26

Desh Raj

8 months

📢📢 **Defending my PhD in a week** Date & Time: January 26, 2024, 9 to 11 AM EST Committee: Sanjeev Khudanpur, Dan Povey, Jinyu Li Dissertation Title: "Listening to multi-talker conversations: Modular and end-to-end perspectives" DM me for a Zoom link if interested 😀

16

7

161

0

0

16

@hbredin

Hervé "pyannote" Bredin

1 year

I am looking for "exotic" audio files to evaluate the robustness of #pyannote pretrained speaker diarization pipelines. They need to come with reliable speaker labels and timestamps, though. Could be a nice way for industry users to give back to #pyannote RT appreciated.

3

16

15

@hbredin

Hervé "pyannote" Bredin

2 years

<bragging> Did I understand the rule correctly? </bragging>

Tweet media one

1

0

13

@hbredin

Hervé "pyannote" Bredin

2 years

@karpathy pyannote-audio to the rescue!

Tweet card media

GitHub - pyannote/pyannote-audio: Neural building blocks for speaker diarization: speech activity...

Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding - GitHub - pyannote/pyannote-audio: Neural build...

3

2

13

@hbredin

Hervé "pyannote" Bredin

9 months

🎉 #pyannote

Tweet media one

@julien_c

Julien Chaumond

9 months

THIS changes everything 🌟

3

3

13

0

2

15

@hbredin

Hervé "pyannote" Bredin

3 years

The proposed approach combines pyannote pretrained speaker segmentation and embedding models hosted on @huggingface hub. It is suitable for real-time applications and its latency can easily be adjusted between 500ms and 5s.

Tweet media one

1

2

13

@hbredin

Hervé "pyannote" Bredin

7 years

Major release with LSTM-based voice activity detection, speaker change detection, and neural speaker embedding.

Tweet card media

GitHub - pyannote/pyannote-audio: Neural building blocks for speaker diarization: speech activity...

Neural building blocks for speaker diarization: speech activity detection, speaker change detection, overlapped speech detection, speaker embedding - GitHub - pyannote/pyannote-audio: Neural build...

0

6

12

@hbredin

Hervé "pyannote" Bredin

3 years

Found a bug 🪳, squashed it 🔨, re-ran the experiments 🔁, and got significantly better voice activity and overlapped speech detection results 🍾 Just in time for camera-ready version 😅 Try it for yourself with this @streamlit demo:

Tweet media one

1

3

11

@hbredin

Hervé "pyannote" Bredin

2 years

What's the "best" pretrained PyTorch speaker embedding models publicly available out there? "best" as in "easy to use" and "with top-notch speaker verification accuracy".

2

0

11

@hbredin

Hervé "pyannote" Bredin

3 years

"Imitation is the sincerest form of flattery" 🤔 github·com/Speech-AI/SpeechX 🤨 🙂

Tweet card media

GitHub - asteroid-team/asteroid: The PyTorch-based audio source separation toolkit for researchers

The PyTorch-based audio source separation toolkit for researchers - asteroid-team/asteroid

1

0

12

@hbredin

Hervé "pyannote" Bredin

1 year

Oyez, oyez, fellow #diarization aficionados! @juanmc2005 will defend his PhD thesis about continual learning and its application to NLP and (online) speaker diarization. When? April 5th, 2PM (CEST). Where? at @LisnLab or online (link below)

2

2

11

@hbredin

Hervé "pyannote" Bredin

1 year

"If you want something done, do it yourself". #pyannote #logo

Tweet media one

2

0

11

@hbredin

Hervé "pyannote" Bredin

1 year

"Open-source is great," they said. Help me reply to this guy and you'll get a free #pyannote sticker in return!

Tweet media one

6

0

12

@hbredin

Hervé "pyannote" Bredin

1 year

The recording of my talk at VoxSRC 2023 workshop is now live. #pyannote #speakerdiarization

@hbredin

Hervé "pyannote" Bredin

1 year

Woot! Woot! #pyannote team ranked third… without ensembling 😇

Tweet media one

1

1

25

0

4

12

@hbredin

Hervé "pyannote" Bredin

4 years

Given the audio recording of a conversation, speaker diarization systems try to answer the question "who speaks when". This is how you can get a (hopefully decent) answer with three lines of code in pyannote․audio:

Tweet media one

1

3

12

@hbredin

Hervé "pyannote" Bredin

3 years

Works from the waveform directly. More details on @huggingface model hub:

Tweet card media

pyannote/embedding · Hugging Face

0

1

11

@hbredin

Hervé "pyannote" Bredin

2 years

Very important addition to pyannote GitHub page!

Tweet media one

2

0

11

@hbredin

Hervé "pyannote" Bredin

5 years

@streamlit is awesome! Here is a screenshot of a demo I built in less than an hour showcasing pyannote-audio pre-trained voice activity detection models. #pytorch #pyannote

Tweet media one

2

2

11

@hbredin

Hervé "pyannote" Bredin

3 years

Happy to share @juanmc2005 's latest work targeting live speaker diarization:

Tweet media one

@ArxivSound

arXiv Sound

3 years

``Overlap-aware low-latency online speaker diarization based on end-to-end local segmentation. (arXiv:2109.06483v1 []),'' Juan M. Coria, Hervé Bredin, Sahar Ghannay, Sophie Rosset,

0

1

5

1

0

11

@hbredin

Hervé "pyannote" Bredin

11 months

(Yet (yet (yet (yet another)))) Whisper + #pyannote tool 💪

@mdj_dev

Maciej Jastrzębski

11 months

🤖🎙️ Exciting update! @Q1w1N and I have been tinkering with AI models lately and today we are sharing as open source a tool for generating meeting transcriptions! It combines: 🧠 Speaker diarization (recognition) 🧠 OpenAI's Whisper automatic speech recognition 💻 User-friendly

2

8

32

2

2

11

@hbredin

Hervé "pyannote" Bredin

1 year

#pyannote "benchmarked" 🤔in marketing announcement of commercial speaker diarization #AchievementUnlocked

@DeepgramAI

Deepgram

1 year

When comparing it to other popular models, we found that Deepgram’s speaker diarization outperforms many commercial diarization models and common #opensource alternatives like #Pyannote when dealing with domain-specific, real-world data.

Tweet media one

1

0

3

4

1

11

@hbredin

Hervé "pyannote" Bredin

1 year

Does that make me a bad person?

Tweet card media

A version of pyannote that functions · Issue #1393 · pyannote/pyannote-audio

Is your feature request related to a problem? Please describe. any attempt to import any pipeline results in a module not found error. All of the pretrained models appear to be tied to a version of...

2

0

10

@hbredin

Hervé "pyannote" Bredin

1 year

2 days to go #Interspeech2023 #pyannote #Anascaul

Tweet media one

1

0

9

@hbredin

Hervé "pyannote" Bredin

2 years

💥 model-assisted annotation recipe with @explosion_ai 's

Tweet media one

1

3

10

@hbredin

Hervé "pyannote" Bredin

11 months

Here is an MVP's MVP 😳 #pyannote #demo #speakerdiarization

Tweet card media

Pretrained pipelines - a Hugging Face Space by pyannote

@maxhbain

Max Bain

11 months

@hbredin is the MVP of diarization. Hard to beat #pyannote (trust me I tried)...

0

0

5

0

1

8

@hbredin

Hervé "pyannote" Bredin

1 year

That was a great success!

Tweet media one

0

0

9

@hbredin

Hervé "pyannote" Bredin

2 years

3/4 <bragging> All top 3 teams used pyannote·audio voice activity and overlapped speech detection </bragging>

1

0

8

@hbredin

Hervé "pyannote" Bredin

5 years

I made a thing! AdaptiveBackgroundGenerator makes your (slow) generator work in the background and automagically adapts the number of workers to keep pace with a (faster) consumer. (1/2)

Tweet media one

2

0

9

@hbredin

Hervé "pyannote" Bredin

1 year

🤫

Tweet media one

@hbredin

Hervé "pyannote" Bredin

1 year

Would you pay for #pyannote custom or premium models? #AskingForAFriend (I'd appreciate a few ReTweets 🙏)

6

27

15

0

1

7

@hbredin

Hervé "pyannote" Bredin

1 year

This morning!

@hbredin

Hervé "pyannote" Bredin

1 year

Alexis Plaquet will be at #Interspeech2023 to present this work. Go talk to him (and grab your #pyannote sticker!)

1

0

7

1

0

7

@hbredin

Hervé "pyannote" Bredin

1 year

Which feature would you like to see added to #pyannote ? See details in thread.

Speaker separation

18

Speaker tracking

9

Multi-channel diarization

5

Streaming diarization

14

3

2

9

@hbredin

Hervé "pyannote" Bredin

8 months

Featuring #pyannote via WhisperX

@tryolabs

Tryolabs

9 months

This year, we're thrilled to introduce some hidden gems, a special selection showcasing a diverse range of innovative and impactful Python libraries. 🌟 Dive into our carefully curated collection. Did we capture your favorite?

0

1

7

1

0

8

@hbredin

Hervé "pyannote" Bredin

1 year

Last day before #Interspeech2023 Free stickers if you find #pyannote #CliffsOfMoher

Tweet media one

2

0

7

@hbredin

Hervé "pyannote" Bredin

2 years

Lessons learned from #VoxSRC 2022 speaker diarization debriefing workshop: a thread.

1

0

8

@hbredin

Hervé "pyannote" Bredin

10 months

6 more months... Keep up the pace! #pyannote

Tweet media one

@hbredin

Hervé "pyannote" Bredin

1 year

6 months later…

Tweet media one

3

0

21

0

0

8

@hbredin

Hervé "pyannote" Bredin

4 years

Just in time for (soon to be cancelled?) #ICASSP2020 , I just released a bunch of pretrained pyannote․audio speaker diarization models and pipelines on @PyTorch torch.hub. Try them out in @GoogleColab :

Tweet card media

Google Colab Notebook

Run, share, and edit Python notebooks

colab.research.google.com

1

2

7

@hbredin

Hervé "pyannote" Bredin

2 years

🤗 pretrained @PyTorch models on @huggingface model hub All models and pipelines were trained on "Jean Zay" french supercomputer. Thanks @Genci_fr ! Find them at

Tweet media one

1

0

8

@hbredin

Hervé "pyannote" Bredin

3 years

Last but not least, @juanmc2005 has a really cool demo based on @ReactiveX RxPy library in the companion @Github repository that can be used to perform live speaker diarization of your microphone!

Tweet card media

GitHub - juanmc2005/diart: A python package to build AI-powered real-time audio applications

A python package to build AI-powered real-time audio applications - juanmc2005/diart

0

0

7

@hbredin

Hervé "pyannote" Bredin

11 months

Woot woot!

@thomas_mol

Thomas Mol

11 months

🚢 Shipped -> improved speaker label editing for , now you can easily select, edit and add new speakers with a dropdown menu ✂-> Manually merge or add new sections to the transcript ⏩-> Added a +15s and -15s playback buttons to easily navigate the audio

0

2

5

0

1

7

@hbredin

Hervé "pyannote" Bredin

2 years

👩‍💻try the online @streamlit demo!

Tweet media one

2

1

7

@hbredin

Hervé "pyannote" Bredin

1 year

#lazytwitter #speakerdiarization What's the current state-of-the-art on DIHARD 3 (full)? I am aware of the winning ensemble submission reaching DER=16.8% but am more interested in single system performance. We're quite bad at tracking these numbers as a community, aren't we?

1

2

7

@hbredin

Hervé "pyannote" Bredin

2 years

😓 (almost) complete rewrite of the codebase 🐍 Python-first API ⚡ faster training with @PyTorchLightnin 🎛️ data augmentation with torch-audiomentations

Tweet media one

1

0

7

@hbredin

Hervé "pyannote" Bredin

1 year

LOTS of ideas @yehor_smoliakov ! Just not on my priority list as #pyannote is first and foremost a way for me to ensure reproduction of my academic research (hence focusing on accuracy rather than efficiency) 👇

@yehor_smoliakov

Yehor Smoliakov

@yehor_smoliakov

1 year

@hbredin Congrats! Any ideas how to make diarization faster in PyAnnote?

0

0

1

1

0

6

@hbredin

Hervé "pyannote" Bredin

1 year

Alexis Plaquet will be at #Interspeech2023 to present this work. Go talk to him (and grab your #pyannote sticker!)

1

0

7

@hbredin

Hervé "pyannote" Bredin

9 months

Looking for examples of audio files where @openai Whisper hallucinates. Would be nice to have some kind of crowd-sourced repository where people can share those examples.

3

0

7

@hbredin

Hervé "pyannote" Bredin

3 years

Been told a few times I should write 📝more about 🎹 pyannote․audio. So I started a blog...

1

0

6