Spanish transcriptions contain text "This is the Spanish subtitles channel of The Church of Jesus Christ of Latter-day Saints"

[ivansabik]

When transcribing multiple videos using Spanish language, I am getting this in multiple videos:

"Este es el canal de subtítulos en español de la Iglesia de Jesucristo de los Santos de los Últimos días"

This translates to something like:

"This is the Spanish subtitles channel of the The Church of Jesus Christ of Latter-day Saints"

I am using a really simple code:

whisper_model = whisper.load_model("medium")
transcription = whisper_model.transcribe(audio_file, language="es")

Examples:

• https://www.youtube.com/watch?v=0KTa6RgCI-I
• https://www.youtube.com/watch?v=RpR0LGUA3zg

Anybody has an idea of how we can solve this or where it could be coming from?

[glangford]

Could be the same underlying issue as

• Dataset bias ("❤️ Translated by Amara.org Community") #928

[ivansabik]

Definitely the same, thanks for pointing to that one