faster-whisper

faster-whisper copied to clipboard

Align segment structure with openai/whisper

1

At the moment, the segments returned by faster-whisper are a subtly different shape to those returned by Whisper and also lack some of the information that JSON/verbose transcription results from...

tekacs

timestamp not matching well when run transcribe on two mono audios split from stereo and assemble back

2

It seems that timestamp from whisper is not very accurate, so when run two mono audios separately, and trying to assemble back the origin audio based on timestamp isn't very...

junchen6072

batch execution transcribe in faster-whisper

38

Hello, Is batch execution of faster-whisper's transcribe possible? We've seen in this [thread](https://github.com/OpenNMT/CTranslate2/issues/1119) that batch execution should increase the throughput. But it's not clear how to perform batch using faster-whisper...

eschmidbauer

Preserving pts time gap on the audio to maintain video/audio sync

2

For use case where whisper is used to transcribe video stream like mp4, sometimes the original stream (e.g RTMP) has network issues that results in pts (presentation timestamp) jumps. Currently...

dodysw2

test short audio file with silence between sentence, word timestamp may not be accurate

4

version: master (e9a082dcf27647eb52585cda6f115454c0ac6856) voice file: [test.mp3.zip](https://github.com/guillaumekln/faster-whisper/files/11170404/test.mp3.zip) voice file: Sentence and sentence duration of silence > 2s audio wav: set enable vad_filter and word_timestamp. Testing different scenarios, the word cannot be...

johnchienbronci

Understanding multiple languages makes unwanted Translations

2

I was using the model (large-v2) for transcribing our meetings. But we were using 2 languages with the speaker 1. native language (Turkish) and the 2. part for checking the...

MustafaCQN

Language specification of 'EN' not working

5

Hello, I am running a two weeks old version of faster-whisper. The thing is that I only need English transcriptions, so I specify it as a parameter (language="en"). It does...

madalinstan-greppysystems

This is very cool, but push to even higher gpu usage?

8

First, thank you for this awesome work and it indeed improves the transcribe time a lot! But I'm wondering if it's possible to push to even higher gpu usage so...

junchen6072

Push model towards outputting shorter segments

1

As part of trying to fix the misaligned timestamps with VAD use I am trying to have the model output shorter segments on average. https://github.com/openai/whisper/discussions/223 There is a solution suggested...

mayeaux

CUDA + fp16 + windows, faster on Large-v1 than on Medium ?

3

Hello, I ran some benchmarks on models on a RTX 2070 and I end up with the following results : ```txt Medium GPU cuda float16 | file 874 seconds :...

ExtReMLapin