Hi ! I tried deepgram model nova 2 + Wait seconds 0.8 + endpointing 300 ms; call id where user said yes (in romanian language 4 times) and it was not transcribed 6ab3a089-7255-411a-9072-5d9746ef1cf4 @User I also added custom keywords via API for yes, two. Still not working

@Gabriela Macovei Hello! I'm sorry to hear that you're experiencing issues with transcription. Unfortunately, Deepgram does not support Romanian language transcription. However, I can recommend using the Talkscriber Whisper model, which may provide better results for Romanian language audio. If you'd like, I can guide you on how to use the Talkscriber Whisper model for your transcription needs. Would you like to proceed with that?

tried. same issue with one words

@Gabriela Macovei looking into it, allow me sometime to do debugging on my side.

after changing to azure, it seems to work better. 80% of the time it understands showr words, but stil not 100%

@Gabriela Macovei Good to hear it's resolved 80% of the time. Since these systems are relatively new, they may not work 100% of the time, but more than 80% reliability is what you can expect. Could you please share the following information: - The recent call ID - When exactly this happened (timestamp) - What response you expected to get - What response you actually received