I have a partner from an asian country that likes subtitles in her native language, but they're not always available. I wrote a PowerShell script that takes an English SRT file and feeds it to GPT4 in chunks to translate. The constraint is actually the 4K token output window because asian languages are not efficiently tokenized, so each word takes 3-4 tokens.
She said the translations are perfect.
The main issue I had was the API profanity and "safety" filter, which kept tripping up on "violence" and "swearing".
I didn't write the script, and I wasn't asking GPT to give me tips for committing murder. The puritan nannying of API users is insane.
Imagine a SQL database refusing to answer a SELECT statement because some text column had potty words in it!
She said the translations are perfect.
The main issue I had was the API profanity and "safety" filter, which kept tripping up on "violence" and "swearing".
I didn't write the script, and I wasn't asking GPT to give me tips for committing murder. The puritan nannying of API users is insane.
Imagine a SQL database refusing to answer a SELECT statement because some text column had potty words in it!