argmaxinc / WhisperKit

On-device Speech Recognition for Apple Silicon
https://takeargmax.com/blog/whisperkit
MIT License
3.17k stars 267 forks source link

Respect skipSpecialTokens option in the decodingCallback function #115

Closed shawiz closed 5 months ago

shawiz commented 5 months ago

I want to show the live transcription process as the transcription happens, and don't want it to show special characters. Right now, regardless of how you set skipSpecialTokens in the decoder options, it will include special characters in the decodingCallback function under .text. I fixed that so it follows the skipSpecialTokens option. I've verified it in the WhisperAX app.

atiorh commented 5 months ago

Thanks @shawiz ! I will let @ZachNagengast chime in. On a separate note, we are currently lowering the "Eager mode" from the app layer down to the library so it can be built on without extra user code. Just mentioning it now, in case you had to rewrite or copy-paste code to use it so far.

shawiz commented 5 months ago

Thanks guys!