We are using the realtime model of chatGPT, the customer complained to us that the AI is not speaking correctly, can you identify the issue and give us some step to fix it.

Please help me to know which other voice services will support ssml tags other than 11 labs.

Hi! So, I'm building a software for scheduling calls for agencies, and I've worked pout my workflow for my entire coding operation, up till the point of making proper phone calls, among other functionalities being configured properly. Upon testing the created outbound calls, the latency seems a bit too much for the outbound calls, its very normal for inbound calls, and web calls however. How do I fix that?

Please help me to know which other voice services will support ssml tags other than 11 labs.

Wanted to clarify 3 things from the docs **1. Use HIPAA-compliant accounts with all third-party providers (STT, LLM, TTS). **Does this just means we must select the right providers (i.e. Deepgram, OpenAI, Vapi) for our assistant, right? Do you handle BAAs with those orgs? I don't want to go wrangle BAAs & update accounts with all of them. 2. Tool Calls / Integrations. How does integrating with other services work with tool calling? We might be interfacing with external providers to accomplish our work over the phone (i.e. we get a patient reference number and pull in specific data for that patient) and need to be able to execute tool calls while on the phone. 3. End of Call workflows. Our server is compliant. We need to get end of call reports w/ transcripts to understand what happened, process the data, and share it with our clients. Is that functionality still supported in a HIPAA-enabled instance?

Dear Vapi Support Team, I would like some assistance on configuring my inbound calls using my own SIP trunk (BYOC) with Vapi.ai to leverage my existing PBX (Yaxxa in this instance in NZ). I have tried following the steps provided in the documentation and have also connected with Yaxxa to no avail. Would love to connect with one of the engineers to know what would be the issue when connecting my own SIP Trunk.

Hi team, I submitted an urgent ticket via email regarding a stuck Telnyx number (+46 72 810 08 03) previously used with ElevenLabs. It's blocking our production go-live. Could someone please take a look? https://cdn.discordapp.com/attachments/1447276476327006421/1447276476629127309/image.png?ex=693708c1&is=6935b741&hm=9dd593cee7e08b4a7b3c2c9fdbda96883db66e5c830ca4c7a0405b8a24492fa1&

Hi! So, I'm building a software for scheduling calls for agencies, and I've worked pout my workflow for my entire coding operation, up till the point of making proper phone calls, among other functionalities being configured properly. Upon testing the created outbound calls, the latency seems a bit too much for the outbound calls, its very normal for inbound calls, and web calls however. How do I fix that?

@Vapi Support Bot @VAPI Tech I recently uploaded a campaign with over 1,000 customer contacts. However, I noticed that only 128 calls were made before the campaign was marked as completed. Could you please clarify why the system did not attempt calls to all 1,000+ customers? I have the auto-reload feature enabled, so whenever credits are close to zero, additional credits are purchased automatically. Given this setup, I expected the campaign to continue until all contacts were reached. I would appreciate your assistance in understanding: - Why the campaign stopped after 128 calls. - Whether there are any settings or limitations that may have caused this. - How I can ensure future campaigns reach all intended contacts. https://cdn.discordapp.com/attachments/1447234941153841333/1447234942030581771/image.png?ex=6936e212&is=69359092&hm=e10341fd1d7e1c3f56c01ead0a56cff26136fcd6724dc06e28a954dcf09601da&

I am building a custom transcriber for Vapi using ElevenLabs Scribe v2 Realtime. The issue is that Vapi sends audio to the custom transcriber that includes both: (1) the customer’s audio, and (2) the assistant’s TTS audio. Since ElevenLabs Scribe v2 Realtime does not support multichannel metadata, diarization, or channel_index in WebSocket mode, all audio is treated as a single mixed PCM stream. This causes the assistant’s own spoken audio to be transcribed and returned to Vapi, which Vapi then interprets as user input. This results in continuous self-interruption loops. I need clarification on the following: 1. Does Vapi send a mixed mono/stereo PCM stream to custom transcribers, or does it send separate channels? 2. If stereo: what is the exact channel mapping (which channel is user, which is assistant)? 3. If mono: is there any way to configure Vapi so that the custom transcriber receives only user audio? 4. Is there any documented or undocumented setting to prevent assistant TTS audio from being forwarded to the transcriber? 5. Is channel metadata supported or planned for custom transcriber mode? This information is necessary because ElevenLabs Scribe v2 Realtime cannot perform channel separation or diarization on a mixed stream, so the current architecture forces incorrect transcriptions.

Hi team, I'm working on creatign an evaluation suite for our assistants and frequently running into issues with the turn based evals, "Expected 2 assistant messages but got 1" or varying. Question: Considering agents are non-deterministic, how can I implement an eval that doesn't fail if the agent says 1 extra line or call tools at different times? Eg. Sometimes the agent will call a tool to verify patient immediately upon receiving all patient details, sometimes the agent may say "let me check your details" then call the tool. Are evals only built for deterministic testing? Should we be using test suites?

Not a single one of the specific voices where you would hit the copy button and paste it in works at all.

It's been out since August. What is going on? https://cdn.discordapp.com/attachments/1446986172684308582/1446986172994682951/Screenshot_2025-12-06_at_4.49.56_PM.png?ex=6935fa63&is=6934a8e3&hm=de1e5931ad08dd6e3fa34bfb71f1671e12ebb4c10f091171317a34826e298990&

No matter what I do to the prompt, it will not change the output. I'm stuck with this demo for "Riley" That it defaults to when you first create an account. "Wellness Partners" demo, I just can't get out of it. I'm trying to use Rime Groq Assembly AI combo. Changing those options doesn't seem to change anything either.

Whenever I try talking to the assistant, I get the error "unable to set up call please try to rejoin." I created another fresh account and I still get the error. Has been like this since yesterday afternoon in midwest US region

It’s not clear whether the test suite is still supported or not. It works and is used very differently compared to Evals, and there’s no clear update in the documentation about whether it can still be used. According to this page https://docs.vapi.ai/test/test-suites is still active, and the API endpoints still seem to work, but they’ve been removed from the API documentation, even though the Vapi UI still allows us to use them. Is the test suite still supported? Why was it removed from the API docs? What’s the plan going forward?

I am creating bulk assistants using APIs and launching them using web calls only do i need to create a new Public key for evey new assistant to be able to use it using public calls , because every time i test I got the error : { "message": "Key doesn't allow assistantId 'd61db6a0-ca3f-4377-9132-0e8a0c7ce0cd'.", "error": "Forbidden", "statusCode": 403 } Unless I create a new public key and allow the newly created assistant

any time I try to tranfer a call to an external phone number from the AI, it just fails. Is the VAPI team looking into this?? seems like everyone is having this problem with no real solution

Despite being visible in the Vapi Voice Library, the voice profile “Neha” (Female, Indian American) does not appear in the search bar when attempting to select it under Assistant → Voice Configuration. When typing “Neha” in the voice selection field, the system returns “No voice found”, even though the voice is clearly listed and available in the main Voice Library panel. This issue prevents the user from assigning the Neha voice to an assistant, causing configuration blockage and inconsistent synchronization between the Voice Library and the Assistant Configuration module. Expected Behavior: Any voice displayed in the Voice Library (including Neha) should be searchable and selectable in the Assistant Voice Configuration dropdown. Actual Behavior: Neha’s voice appears in the Voice Library but is not recognized or retrievable in the assistant’s voice selection search bar. Impact: Blocks the user from using the selected voice for production. Creates confusion due to mismatch between library visibility and configuration availability. Interrupts the workflow of creating Hindi or Indian-accent voice agents. https://cdn.discordapp.com/attachments/1446800988039155824/1446800989561815040/Screenshot_2025-12-06_151524rg.png?ex=69354dec&is=6933fc6c&hm=ef10c8173d7fdb2a6419c04296714f222ba62254e393040e269dddac612ba7de& https://cdn.discordapp.com/attachments/1446800988039155824/1446800990320857168/Screenshot_2025-12-06_151549Sd.png?ex=69354dec&is=6933fc6c&hm=593ce02348a01c180605b7c6a22991b81bdae170c80a412c8ce68a202213a1be&

The call is not triggered even though the schedulePlan is set as follows. How can I resolve this issue? "schedulePlan": { "earliestAt": "2025-12-06T10:05:00Z", "latestAt": "2025-12-06T10:11:00Z" }

The error message says "Fix validation errors to publish" but im not sure where that validation error is at. there is no error indication somwhere to follow. this only happends when i try to publish the gpt-realtime clusters. https://cdn.discordapp.com/attachments/1446787121099706451/1446787121569337395/Screenshot_1.png?ex=69354101&is=6933ef81&hm=c37da3fa5001b6191909271ac362dbb601ed82616d9bfd78df4672399128cd33&

I keep getting this error , I checked the other thread and this problem havent solved . Can someone please help ? https://cdn.discordapp.com/attachments/1446744963080781955/1446744963432972519/Screenshot_2025-12-06_010329.png?ex=693519be&is=6933c83e&hm=9b3c9c23176b90a07996beff9c839944857b31bcb327f0494535fe792b3884e8& https://cdn.discordapp.com/attachments/1446744963080781955/1446744963873378364/Screenshot_2025-12-06_010641.png?ex=693519be&is=6933c83e&hm=f27b47ce72f3d0e1992d26677c7dbc3106d7b5f9c0281f201239b624bbf50a03&

Reading through the documentation, there doesn't seem to be a way to "end" a chat so that an end of call report is generated to get the structured data.

I need the receptionist to first check with the tool/n8n webhook to make sure the person's phone number is on the list. then i need it to run the compliance script and get confirmation they agree to recording, then make the transfer. the end of call webhook is configured on the assistant the call is being transferred to. Thanks in advance. The json for both workflows is attached if you need to look at it. the latest call log is: 019aec28-2b51-7113-a0ea-039d43205900 squad link: https://dashboard.vapi.ai/squads-v2/bad2ef63-78a0-4d4e-8a4d-31bafcf300cf assistant link: https://vapi.ai?demo=true&shareKey=4fe97633-2e81-4493-9feb-8a34592767cc&assistantId=bc7aa79c-e0af-42e4-b5aa-1e73489be085

Not able to Connect with GoHighLevel

Details here https://github.com/VapiAI/client-sdk-react/issues/35

When I'm making Campaign call, It's not using all details in csv. It supposed to use Name and all other info I've added in csv data for campaign. Instead it's only making call to phone number. Even when I'm trying to make API Call, still it's not using all data that we have.

I'm having a problem with my agent. It's entirely in Brazilian Portuguese, using a native voice from 11labs. When I call it, the voice is completely buggy, with a strange accent, and unreadable. However, when I use VAPI's "Talk to Assistant," the voice is perfect, with no accent or speech errors. Can anyone help me? Do I need to make any specific settings?

This issue have been happening a while. What is the main problem related to it? https://cdn.discordapp.com/attachments/1446563511508402414/1446563512234020946/image.png?ex=693470c1&is=69331f41&hm=783d239cca8abe1d60e8ec8b00696b85e82efa88d8e05cdd3e4312a7bedaf460&

I’m trying to start a call via the Vapi API, and I’m getting this error: "Couldn’t get tool for hook.

toolId

a4e835ca-b846-4b23-b748-c1c7a1ed4e39 does not exist." However, the assistant I’m trying to call has no tools assigned to it. My orgID: 61d09b13-129f-4353-9eb3-0e7ee89eb6b8 Assistant ID: 378bbdc9-e255-46ca-b6b9-a6ce06c18961