Lightning Whisper MLX

Posted by mustafaaljadery on Tuesday April 02, 2024 @12:24PM
from the github.com dept.

( 6 comments | 9 points | on HN )

Re:Lightning Whisper MLX by siraben on Tuesday April 02, 2024 @03:41PM (#39910028)

What's the word error rate? Is it the same as the distilled whisper models?

[ Parent | on HN ]

Re:Lightning Whisper MLX by mustafaaljadery on Tuesday April 02, 2024 @12:24PM (#39907568)

An extremely fast implementation of whisper built on top of Apple MLX. It outperforms Whisper CPP by ~10x.

[ Parent | on HN ]

Re:Lightning Whisper MLX by jxy on Tuesday April 02, 2024 @04:53PM (#39910762)

How did you run whisper.cpp? Can you share the command line? What machine did you use to run that benchmark figure?

[ Parent | on HN ]

Re:Lightning Whisper MLX by asdlkj178 on Tuesday April 02, 2024 @11:40PM (#39913386)

Will disable beam search hurt performance on long form transcription?

According to the whisper paper, beam search is necessary to achieve good long form audio transcription performance.

[ Parent | on HN ]

Re:Lightning Whisper MLX by kanwisher on Wednesday April 03, 2024 @03:54AM (#39914680)

Are you going to have a Swift version of this?

[ Parent | on HN ]

Re:Lightning Whisper MLX by sebnun on Wednesday April 03, 2024 @11:56AM (#39919097)

Nice. Why was this tested with a 15 min video instead of an audio file? or is it a typo on the graph?

[ Parent | on HN ]