Hello 🙂 I am attempting to setup a Welsh speaking chatbot. My prompt is written in Welsh, using GPT-4o model. For transcription I have Azure set to cy-gb (their welsh model). For voice I have Azure cy-GB-AledNeural, their male welsh voice. It appears to produce audio that is welsh, but in the textual transcript it is completely nonsensical. Call ID: 904feeab-dd96-49fa-a3f5-c4de7186e52c

HeyTopsyMustang Thank you for creating the ticket! A member of our team will reach out to you shortly.

@TopsyMustang Hello! It's great to hear that you're setting up a Welsh-speaking chatbot. I understand that you're experiencing issues with the textual transcript of the Welsh audio. This could be due to a variety of factors, including the complexity of the Welsh language for speech recognition systems, or a mismatch in the configuration. To assist you better, I can help you review the settings and ensure that everything is configured correctly for both transcription and voice synthesis. Could you please confirm if the settings you've mentioned (Azure set to cy-gb for transcription and Azure cy-GB-AledNeural for voice) are currently active for the call ID you've provided? If there are any specific settings or parameters you'd like to adjust or verify, please let me know, and I'll be happy to assist you further.

@Vapi the settings i can confirm are currently active

@TopsyMustang Great, thank you for confirming that the settings are currently active. It sounds like the configuration is correct, but the transcription quality is not meeting expectations. This could be due to various reasons, such as the quality of the audio input, the accent or dialect of the speaker, or limitations in the speech recognition model itself. To proceed, I would recommend the following steps: 1. **Check Audio Quality**: Ensure that the audio input is clear and free from background noise. 2. **Speaker's Accent/Dialect**: Confirm if the speaker's accent or dialect closely matches the standard that the Azure model is trained on. 3. **Model Limitations**: It's possible that the Azure speech recognition model may have limitations understanding certain words or phrases in Welsh. If after checking these factors the issue persists, it might be necessary to reach out to Azure support for a more in-depth analysis of the problem. They might be able to provide insights into whether the model has known limitations or if there are any updates or additional configurations that can improve transcription accuracy. If you need further assistance with this or any other settings, please let me know!

I need further assistance please

Hey @TopsyMustang To help track down this issue, could you share: - What response you expected to get - What response you actually got instead This would really help me figure out what went wrong!

@Shubham Bajaj sorry just putting together the relevant info