Max_line_width doesn't work for me

[petqo]

Hello,

I use whisperAI in google colaboratory, works fine and fast but for some reason --max_line_width doesn't work, does anyone has this same problem?

here's the command I use

!whisper untitled.mp3 --language cs --model medium --word_timestamps True --max_line_width 16

[ryanheise]

Can you copy/paste part of the generated srt file to illustrate the problem?

[ryanheise]

For comparison, can you paste the output with --max_line_width 32?

[petqo]

1
00:00:01,500 --> 00:00:09,820
Hrubé znamení, které se snaží
odradit každého, kdo by chtěl
proniknout hlouběji do jeskyně.
Držte se dál, nebezpečí uvnitř,
zákaz vstupu.

2
00:00:10,580 --> 00:00:12,780
Dobře, jdu tam, co nejhoršího se
může stát.

3
00:00:13,580 --> 00:00:24,280
Ty pičo! Co ti jebe! Hoši! Hoši!
Hoši! Hoši! To si děláte prdel!

[petqo]

Oh so it's not like the lines matches the timestamps and makes "shorter" subtitles... My "goal" is to make shorter lines with timestamps, cause I want to use that for tiktok videos, where you give only 1-4 words to the screen

[ryanheise]

You can use --max_line_width in conjunction with --max_line_count. The former decides when to start a new line, and the latter decides when to start a new subtitle. Note that the latter overrides Whisper's native intuition on when to start a new subtitle which often (but not always) aligns with phrase and sentence boundaries, so by overriding that you will just get it starting a new subtitle whenever the length limit has been reached. There is one exception where it will try to start a new subtitle earlier than the length limit if there has been a pause of a few seconds.

[petqo]

thank you so much!!

[wenchaoliu-93]

You can use --max_line_width in conjunction with --max_line_count. The former decides when to start a new line, and the latter decides when to start a new subtitle. Note that the latter overrides Whisper's native intuition on when to start a new subtitle which often (but not always) aligns with phrase and sentence boundaries, so by overriding that you will just get it starting a new subtitle whenever the length limit has been reached. There is one exception where it will try to start a new subtitle earlier than the length limit if there has been a pause of a few seconds.

It doesn't work for me.
Command: time whisper He\ Is\ Risen.mp4 --model tiny --output_format srt --max_line_count 1 --max_line_width 42 --word_timestamps True --device cpu

Sample Output: [00:02.000 --> 00:14.360] This has been a remarkable session, and behalf of all who participated thus far in Word or music. [00:16.580 --> 00:22.860] As the President of the Church, I've chosen simply to say to you on this moment. [00:24.700 --> 00:30.780] And just two words, known as the two most important words in the English language,

[wenchaoliu-93]

Sorry. It looks like the whisper output on the terminal can be different from the actual output file.