openai/whisper-large-v3-turbo · Model incapabilities in Turkish

Hi, I am testing this model on call center calls but it is incapable of capturing every conversation. Compared to the 4o-mini-transcribe API, this model is relatively bad to capture and transcribe the whole conversation.
The sentences like "Thank you for watching", "Subtitle by .." etc. also being put instead of actual conversation on that part. So that part of the conversation there is lost. This not happen in 4o-mini-transcribe.
The last issue is not being able to provide an initial-prompt. There are keywords like terms or special names of companies, cities etc. that I want it to capture but cannot provide a word list. There is an option to pass a prompt in 4o API models.

Do you also have similar issues? If you have can you enlighten me on how to solve these kinds of issues? I do not want to use the 4o API since it is a bit costly.