Lexicaps - Whisper Diarization - no third-paryt packages

[Majdoddin]

www.lexicaps.com

Augmenting Whisper's transcription by adding "speaker tags", so you know who says what. Currently it works for 2 speakers, and is tested for English.
I trained a classifier on top of Whisper model features (medium.en), that identifies any two speakers. No third-party package is used for Diarization.
Seamlessly integrated with Whisper, it provides a comprehensive Transcription-Diarization service.
Give it a try or show a Sample.
Discussions: Majdoddin/lexicaps

Thanks @karpathy for the fun project
Thanks @sidhantls for inspirative repo

[databill86]

Great work!
Are you planning to publish the source code ? :)

[ChipNowacek]

lexicaps.com -> "We are maintaining our server. We will be back soon."
https://github.com/openai/whisper/discussions/github.com/Majdoddin/lexicaps -> 404 (seems reasonable given the URL)
https://github.com/Majdoddin/lexicaps -> just what's on this page (above)

Throw me a bone?

[freshpearYoon]

hi, thanks for great work! Is there any plan to publish source code?

[sfc-gh-praj]

@Majdoddin : this is very much needed, please let me know if you are planning to publish your code..