Cerlancism/chatgpt-subtitle-translator

Moderation warnings among the text

Closed this issue · 1 comments

romanr commented

Here's the output fragment:

445 "a bad man. You are
unworthy, you are stupid, you are" -> "[Flagged][Moderator] a bad man. You are
unworthy, you are stupid, you are -> harassment: 0.494 "
[CheckModeration] flagged [ { catergory: 'harassment', value: 0.9986397624015808 } ]
446 "ugly, you are failure, you"

Can moderation be turned off?

Add the command args

--no-use-moderator

The purpose of the moderator is to check the likelihood if the model itself is unwilling to translate, and narrows from a large batch of 100 lines by default to the specific lines. If you turn the moderator off and the model is refusing to translate, it might tag it as [Flagged][Model]