Good day all,
I am trying to transcribe audio from conversations we have with our customers. Most of the conversations is in Afrikaans. I have tried a bunch of AI transcription services but so far they have not been great. I currently use OpenAI's Whisper model to convert the audio to text from Afrikaans and then translating to English, but the results are below average.
My current settings for Whisper:
Model: medium (multi lingual)
Language: Afrikaans
Translate to English
If I do not translate to English, 50% of the text is a mix between Dutch and German, which is not usable. By translating to English you can at least read the output.
Have anyone tried the above with any language other than English with okayish results? We do not need 100% accuracy, only good enough to search for specific conversations and summarise using AI.
Thanks!
EDIT: Found this issue on github: https://github.com/openai/whisper/discussions/872 . Looks like Afrikaans is supported yes, but not accurate at all.
I am trying to transcribe audio from conversations we have with our customers. Most of the conversations is in Afrikaans. I have tried a bunch of AI transcription services but so far they have not been great. I currently use OpenAI's Whisper model to convert the audio to text from Afrikaans and then translating to English, but the results are below average.
My current settings for Whisper:
Model: medium (multi lingual)
Language: Afrikaans
Translate to English
If I do not translate to English, 50% of the text is a mix between Dutch and German, which is not usable. By translating to English you can at least read the output.
Have anyone tried the above with any language other than English with okayish results? We do not need 100% accuracy, only good enough to search for specific conversations and summarise using AI.
Thanks!
EDIT: Found this issue on github: https://github.com/openai/whisper/discussions/872 . Looks like Afrikaans is supported yes, but not accurate at all.
Last edited: