Whisper parameter
charlo06 opened this issue · 4 comments
Hey. First of all very nice work on the new whisper feature. Unfortunately you can only input the audio file to the whisper api. But the api reference has several other very useful inputs, such as the language and the prompt inputs. Do you have any plans to add them in the future? The prompt feature is really useful to specify certain words like the spelling of unusual names or invented words. It can also be use as a way to specify the language the model is about to heard to correctly transcribe ( a problem that can occur when the speaker has an accent).
Anyway thank you again for your great work. BTW I'm still trying for that stream feature but I guess I'm not good enough in c++ for that 😅 (not giving up though)
nice! I'll work on adding the other parameters listed in the api reference this weekend.
You're the best!
quick update: Almost done adding the other whisper parameters. Travelling to Japan for a couple weeks. I'll wrap it up right when I get back.
Hi, were the extra Whisper parameters ever added? I could really use the prompt to mitigate the random words in response to blank audio.