@Felix WIlliam Hello! It's great to hear that you're using VAPI's platform for your real estate agency in Bucharest. I understand the frustration with the transcriber not recognizing short, one-word Romanian answers. While the Nova 2 model and language settings are currently set to English, which might be contributing to the issue, there are a few steps we can take to improve recognition of Romanian responses: 1. **Language Configuration**: Ensure that the transcriber is set to the correct language. If the Nova 2 model supports Romanian, we should configure the transcriber to use Romanian instead of English. 2. **Model Optimization**: If Deepgram provides a model that is more optimized for Romanian or short utterances, we should consider switching to that model. 3. **Acoustic Model Training**: Some speech-to-text providers offer custom acoustic model training. This involves providing samples of speech that the system can learn from to improve accuracy. 4. **Adjusting Sensitivity**: Tweaking the sensitivity settings of the transcriber might help in picking up shorter utterances. 5. **Noise Reduction**: Ensure that the recording environment is as quiet as possible to reduce the chance of the transcriber missing short words due to background noise. 6. **Post-Processing**: Implementing a post-processing step that checks for very short durations of speech and handles them differently might also be a solution. If you haven't already set the transcriber to Romanian and you'd like to proceed with that, please provide the necessary details so I can assist you further. If the issue persists or you need more specialized assistance, I recommend reaching out to Deepgram's support for guidance on model training or additional settings that might be specific to Romanian language recognition. Would you like to proceed with setting the transcriber to Romanian, if supported, or do you need assistance with something else?

My transcriber is already set to Romanian and the problem is not resolved. https://cdn.discordapp.com/attachments/1326210902273560597/1326296904203767880/Screenshot_2025-01-07_at_17.33.11.png?ex=677ee9bf&is=677d983f&hm=d2b215f423d09fd45a947a2064a207d9d5293e695c2ebb3a24818cd346d9df54&

@Felix WIlliam I'm sorry to hear that the issue persists even with the transcriber set to Romanian. Here are a few additional steps we can take: 1. **Keywords**: Since you're expecting specific one-word answers like "imobiliare," you can add these as keywords in the transcriber settings. This can help the transcriber prioritize and better recognize these terms. 2. **Adjusting Confidence Thresholds**: Lowering the confidence threshold for recognition might help in capturing shorter responses, although it may also increase the risk of false positives. 3. **Feedback Loop**: Implement a feedback loop where the system learns from the missed recognitions. By flagging the missed one-word answers, you can potentially improve the model's accuracy over time. 4. **Contacting Deepgram Support**: Since you've tried the recommended settings and the issue still occurs, reaching out to Deepgram support for more advanced troubleshooting or to inquire about custom model training might be necessary. 5. **Alternative Solutions**: If the issue cannot be resolved with Deepgram, you might need to consider using a different transcriber for Romanian language that has proven to be more reliable for your use case. If you'd like, I can assist you with adding specific keywords to the transcriber settings. Could you provide me with a list of one-word answers that you'd like to prioritize for better recognition?

Lets try with this list of one word answers to prioritize: olx, imobiliare, cash, credit

@Felix WIlliam Great, prioritizing these specific one-word answers can indeed help improve recognition. However, as a support bot for VAPI, I don't have the capability to directly modify the settings of your transcriber. You would need to access the transcriber's settings or configuration file where you can specify these keywords for prioritization. Here's what you can do: 1. **Access the Transcriber Settings**: Go to the settings or configuration section of your transcriber. 2. **Add Keywords**: Look for an option to add or prioritize keywords. This might be under advanced settings or a similar section. 3. **Update and Test**: After adding "olx," "imobiliare," "cash," and "credit" to the list of prioritized keywords, save the changes and conduct tests to see if there's an improvement in recognition. If you're unsure of how to do this or if the transcriber does not provide an option to prioritize keywords, please reach out to the support team for your specific transcriber for further assistance. They should be able to guide you through the process or provide a solution tailored to your needs. Would you like to proceed with reaching out to the transcriber's support team, or is there anything else I can assist you with?

Where i acces "Access the Transcriber Settings" ?

@Felix WIlliam To access the transcriber settings, you would typically do this through the platform or service where the transcriber is integrated. Since you are using VAPI's platform, you would access the settings through the VAPI interface or the configuration files where you set up the transcriber. Here's a general guide on how to access the transcriber settings: 1. **VAPI Dashboard**: Log in to your VAPI dashboard where you manage your voice applications. 2. **Project Settings**: Navigate to the project or application where you have set up the transcriber. 3. **Transcriber Configuration**: Look for a section related to voice recognition, speech-to-text, or transcriber settings. This is where you can typically find options to adjust the transcriber's behavior. 4. **Keywords or Advanced Settings**: Within the transcriber settings, look for an area where you can add keywords or access advanced settings. This is where you would add "olx," "imobiliare," "cash," and "credit" to the list of prioritized keywords. If you are unable to find these settings or if the process is different for your specific setup, please consult the documentation provided by VAPI or contact their support team for detailed instructions on how to access and modify the transcriber settings. If you need further assistance or have any other questions, feel free to ask.

Hey @Felix WIlliam To help track down this issue, could you share: - The call ID - When exactly this happened (the timestamp) This would really help me figure out what went wrong!

@Felix WIlliam hi Felix " how did you solve the romanian one word issue ? I have these settings and it is still not transcribing one word answers https://cdn.discordapp.com/attachments/1326210902273560597/1328732660020809789/image.png?ex=6787c638&is=678674b8&hm=30eaddde5ec064692aa0b50851064464d25f4d38feb7c3bf4f46a88caebf90a1& https://cdn.discordapp.com/attachments/1326210902273560597/1328732660381651085/image.png?ex=6787c638&is=678674b8&hm=021abde877c1d8c7c78d0493b84f41b1ce29d00db6bac6d0d265125b3790b437& https://cdn.discordapp.com/attachments/1326210902273560597/1328732660805271613/image.png?ex=6787c638&is=678674b8&hm=c3988d66b806eb898c9bf49d4aa3d91da375554c4a8b278580289b9876f7070c& https://cdn.discordapp.com/attachments/1326210902273560597/1328732661262319726/image.png?ex=6787c639&is=678674b9&hm=e4872b2ccf2e597fee58a5662b9833ff2c23295dfccf727fcbd6781336291077&

Talkscriber is not good. Use “azure” is the best for romanian. This will resolve all the problems

thanks. indeed it word better than deepgram, but still not 100% 😦 it gets short word right from the very first try for about 80% of the words