https://vapi.ai/ logo
Join Discord
Powered by
Wrong Output from the agent
# support
m
Hello, I run some tests with a german voice caller. And at the end of the call he asks me for some appointment suggestions. He wants to say "vielleicht" the german word for "maybe" . But he say "ein Like" which sound similar to the german word, but as a german you would get confused by this. Is there any fix for that? call id: c433b29c-718b-45e4-bc39-e7ff21d79335 timestamp: 1 minute and 18 seconds - 1 minute 25 seconds
v
Hi @Marcel, I'm an AI assistant for Vapi. I'll analyze available sources and get back to you shortly.
----------------------------- **While I generate an answer, here are some sources that might be helpful**: * [Introduction]() * [Web Calling]() * [Orchestration Models]() * [Introduction]() -----------------------------
### Answer It sounds like you're experiencing an issue with the text-to-speech (TTS) component, where the pronunciation of words isn't aligning with the expected output. This could be due to the specific voice or model settings within the text-to-speech provider you are using. Here's what you can do: 1. **Check the Voice Configuration**: Ensure that the voice and the configuration you're using are optimized for the German language. This includes checking if the voice provider and the voice ID are set correctly for German. 2. **Customize or Switch the Voice Provider**: If the current voice configuration is not producing the desired results, consider switching to another voice provider that has better support for German, or customize the voice settings if your current provider allows it. You might need to use a different voice model or adjust specific settings related to language and locale. 3. **Adjustments and Testing**: After making adjustments, perform thorough testing to ensure that the pronunciation and the speech output align with your expectations. This might involve iterative adjustments to find the optimal setup. Here is a basic example of how you might configure or adjust your assistant's voice setting in JSON format:
Copy code
json
{
  "voice": {
    "provider": "desired-provider",
    "voiceId": "specific-voice-id-for-German"
  }
}
Make sure to replace 
"desired-provider"
and 
"specific-voice-id-for-German"
with the actual provider and voice ID you wish to use. For more detailed guidance on managing and configuring voices, please refer to the documentation on [Custom Voice]() [(1)]() or the [Voice Providers]() where you can find specific settings and options available for different languages and dialects.
Got it. Tagging @User for additional help.
m
I switched different voice providers and he still says the same. So I dont think there is the issue. He just says thaht what is written in the transcript.
s
Use 11labs multilingual, gpt-4o. It will fix your issue.
m
changed it, but same issue again call-id: 46de2f44-34aa-408a-bbb1-b0bf2240cf1d timestamp: 1 minute 12 secons to 1 minute 18 seconds
s
I don't know German, so can you please provide me with some text that was spoken by AI, which should be different from the transcript itself?
Last Message by Assistant was Das ist vollkommen in Ordnung, ich werde mich dann nochmals bei Ihnen melden, um den Termin zu bestätigen. Ich danke Ihnen für Ihre Zeit und wünsche Ihnen einen schönen Tag. Auf Wiederhören.
m
He should say: Wann würde es Ihnen passen? Könnten Sie mir vielleicht ein paar Terminvorschläge machen? Which means: When would it be convenient for you? Could you give me some appointment suggestions? But he said: Wann würde es Ihnen passen? Könnten Sie mir ein Likepaar Terminvorschläge machen? the "likepaar" is wrong here, but it is written down in the transcribe
@Sahil is the information enough? Or do you nedd more from me?
s
No, just give me some time. I will look into it.
Let's do one thing. Let's schedule a meeting and see how I can fix your issue.
m
Sorry I wasnt in the office for a few days. Where can I schedule a meeting?
found it
4 Views
PreviousNext
Powered by