Subtitles created by amara.org, QTSS etc

API
1

Many times into SRT appears strange lines:
“Sottotitoli e revisione a cura di QTSS”
“Subs by www.zeoranger.co.uk Diolch yn fawr iawn am wylio’r fideo.”
“Sottotitoli creati dalla comunità Amara.org”
“Sottotitoli a cura di QTSS”
“Sous-titres réalisés par la communauté d’Amara.org”
“info un libro pubblico su www.mesmerism.info”
etc.

510
00:38:51,080 --> 00:38:53,080
Sottotitoli e revisione a cura di QTSS

511
00:39:21,080 --> 00:39:23,080
Sottotitoli e revisione a cura di QTSS

512
00:39:51,080 --> 00:39:53,080
Sottotitoli e revisione a cura di QTSS

in this case, this line appear from 510 to 797… 288 times :face_with_raised_eyebrow:
there is a way to block this lines?

3 Likes
2

Hello everyone, I’ve also encountered the same issue with repetitive lines in transciptions API mentioning QTSS and Amara org. Has anyone found a solution to block these automatic additions?

1 Like
3

Same problem here. But in interactive mode.

1 Like
4

This is the OpenAI developer community forum,

If you need help using non-OpenAI services, I’ll recommend contacting them directly.

1 Like
5

Hi N2U,
why whisper is a non-OpenAI service?

6

You didn’t mention that you where using whisper, leading to some confusion. (Amara.org is a service that makes subtitles) :laughing:

7

Whisper may hallucinate on empty, or almost empty sections of audio, to prevent this you will need to trim away the sections of your audio that doesn’t contain speech.

Another solution is to remove the text you don’t want in your transcript afterwards, it’s not a perfect solution, but it does work. Here’s a list of common hallucinations from GitHub

8

Tnx N2U
I’ll use this to filter hallucinations.

1 Like
9

Your welcome! Always happy to help :heart: