Please have a look at my original post, your chat bot locked and marked it as Solved. But its not.

@stefan I'm sorry to hear that your issue wasn't resolved. Could you please provide me with more details about your original post and the problem you're experiencing? I'm here to help you get it sorted out.

Can a human look in to this? Your chat bot already tried and failed.

can you edit this ticket with your issue and tag me here again.

@Shubham Bajaj When the customer gives a one word answer, its not recognized, except for the fist time. Swedish language only! AI will ask (in Swedish) Is this Matias I'm speaking to? If customer replies: "Yes" nothing will happen and the call will eventually time out. As in this example: Call ID: 95997d6d-3ce5-467b-bc33-0b28a12a7bfd If the customer however replies: "Yes it is" It works as it should. However it works in Swedish if the one worded reply is after the "First Message" in the prompt. From the log: Assistant Hej, jag heter Joakim och ringer från Bra hjälp, vi som hjälper utsatta barn. User Hej ** ** one word reply recognized * Assistant Är det Mattias jag pratar med? * I replied Yes, but not recognized. * Can you also let me know what to write in the "path id string" and where to get it. See picture. Thanks Stefan https://cdn.discordapp.com/attachments/1282532613659295814/1282894457217355879/vapi3.jpg?ex=66e10411&is=66dfb291&hm=496c3ee0c95b2126b37653b859a2541daacb1d036667358d31e3257595b92591&

Hi, I'm an Vapi user here. I detected the same problem with my calls. I'm openning an new thread about it. About your question: The id can be found in the assistant page, top of page, below the assistant name.

hey @stefan can you check call id again there was no response from user after assistant voiced out

Is this Mattias I’m speaking with?

@Shubham Bajaj @Leobaldo Alcantara As I stated in my post, there was as response. I replied Yes, but not recognized/transcribed. This is not just a one time fail. I have tried MANY times, it does not work. Seems like its not just for me as Leobaldo is reporting the same issue. Thanks Leobaldo for your input!

cam you share call ids where it's captured your input(as transcript) even in call recording i can help.

@Shubham Bajaj Im sorry but are you reading my posts at all? As I wrote in my first message on September 8, It works after the "first message" I also provided a copy of the logs. Here it is again. From the log: Assistant Hej, jag heter Joakim och ringer från Bra hjälp, vi som hjälper utsatta barn. User Hej * * *********** **one word reply recognized ************** Assistant Är det Mattias jag pratar med? ** I replied Yes, but not recognized. ******** Same call ID as provided previously: 95997d6d-3ce5-467b-bc33-0b28a12a7bfd

you right after this 🔵 10:20:24:994

assistant

Final Transcript : Är det Mattias jag pratar med?: 0.9577637 your words were not captured, can you try w/o bg noise because your words post capturing were kinda removed/filtered.

@Shubham Bajaj I tried many times before, it was not a one time fail. But have since modified the assistant and for the moment it is working. I do have three other issues. 1. I tried to use Its not doing anything. I have enabled ssml parsing via "Update assistant" Call ID: ee3ac508-a34e-451f-a551-e4f62b8a1412 2. I tried to use emotions, but the same issue. It doesn't matter what i write, no change in the voice. https://elevenlabs.io/docs/speech-synthesis/prompting 3. Im using a Swedish voice, (have tried with several, even made my own clone at elevenlabs.) In every single call the voice changes dialect after a while, it sounds like its a foreigner trying to speak Swedish. Some words are to the point it cant be recognized. Thanks!

Regarding your mentioned issues: [1] we have sent the same to 11labs what you have given to us 🔵 11:50:55:930 Voice Input Formatted: "vi som hjälper utsatta barn. Anledningen till att jag ringer är att snart är sommaren slut.", Original: "vi som hjälper utsatta barn.. Anledningen till att jag ringer är att snart är sommaren slut." 🔵 11:50:55:930 ElevenLabs (Websocket #0) Pushing 110... "vi som hjälper utsatta barn. Anledningen till att jag ringer är att snart är sommaren slut." 🔵 11:50:55:930 [user LOG] Voice input: vi som hjälper utsatta barn. Anledningen till att jag ringer är att snart är sommaren slut. [2] regarding emotionRecognitionEnabled i will say it's not perfect as currently it uses the transcription and llm model output so chances you may not get consistent results [3] when we stream voices during calls, the difference is noticable because of streaming and meeting latency SLA. what I can suggest is to use default values for better consistency. "model": "eleven_turbo_v2", "style": 0.9, //upto your requirement and expectations. "voiceId": "wIdBMZsynaZq6gk0sjXJ", "provider": "11labs", "stability": 0.5, "similarityBoost": 0.75, "useSpeakerBoost": false, "enableSsmlParsing": true, "fillerInjectionEnabled": false, "optimizeStreamingLatency": 2

@Shubham Bajaj I've been doing additional testing and am experiencing significant delays in response times. When I test your assistant on www.vapi.ai I consistently get excellent response times—between 500 ms and 1000 ms at most. However, the two calls below are showing such delays that it's becoming unusable. Please look into this issue as soon as possible. 1500 to 4000 ms is just not cutting it. This call for example: 82770825-0962-473f-adb2-de1ac688feaf Customer says Hello 4 seconds of silence! Customer says Hello again 1.5 seconds of silence Bot finally starts to speak….. Another example: 25c5cebf-5c73-4b60-989d-2420e77ed6d9 Customer says his name. 1.5 Seconds silence Bot: asking for a specific person. Customer: you have reached the wrong person 4.5 seconds of silence Bot finally starts to speak….. I made a new assistant (your preconfigured AVA), standard settings except for gpt4o. Its a bit better, but still not as good as the respons times at www.vapi.ai b304e05c-b8ab-4e24-8652-d3f2dcf8640e Been testing some more, the delay seems get a lot worse when using non English languages. I would really appreciate a solution to this, even if it means having to upgrade to Enterprise/custom plan etc.

for the second call id

25c5cebf-5c73-4b60-989d-2420e77ed6d9

the user input was > Nej, då har det ringt fel. Har kommit till Anders. Lindblad. If you notice the user has spoken in 3 utterances because of this you observed a delay in response.

@Shubham Bajaj Hello again, I just made a new test call. 65f9186d-00dd-436c-81eb-5ebd04f89904 It took the bot 4 seconds to reply after I answered with my name. After said "its me" I waited 8 seconds, no reply from the bot. After I said "hello its me" it took 2.5 seconds before the bot replied. Sorry, but not really usable in the real world. Is there anything I can do to get it down to sub 1 second as it is on your own test bot? Thanks!

You said 00-02 secs: Jag Bot replied started from 06 secs (latency of 03 secs): Hej Jag heter David, Jag söker Michael 🔵 09:52:16:600

user

Final Transcript : Jag: 0.7084961 🔵 09:52:17:718 Getting Speech For

11labs:wIdBMZsynaZq6gk0sjXJ:1:0:0.9:false:4:eleven_turbo_v2_5:16000:be1c2ae015703d72ac7e897280c69e20eae48cf12acb8666614c52b49cc4e794

, "39" Text: "Hej Jag heter David, Jag söker Michael" As visible from logs the llm generated response in require time, voicing out takes time.