eager-peach•3mo ago
VAPI access?
This is what ChatGPT tells me about why I'm struggling with getting voice/audio work. Is it correct?: "You’re using the new version of Voiceflow, where all project templates are based on the chat assistant framework, and voice-specific channels like VAPI or Twilio are no longer automatically available by default.
So even if you choose Lead Qualification, it's still a chat-first agent, unless you've been granted access to voice-enabled deployment channels."
2 Replies
Here’s the short version:
• Voiceflow still supports voice agents – nothing was “removed”.
• New templates are chat-first, but you can turn any agent into a phone/voice bot by connecting the Telephony (Twilio) integration or by using the Dialog Manager API with your own audio front-end.
• No special beta or private access is required; you just need a Twilio account (a free trial works) and Editor rights in the workspace. What actually changed ───────────────────── The current builder uses a single “Agent” model. We ship most templates with a web-chat preview because that’s the fastest way to test. Voice is now treated as a deployment channel you add afterwards: 1. Open the agent ➜ Integrations ➜ Telephony.
2. Import or buy a Twilio phone number and save the SID/Auth Token.
3. Assign the number to your Development or Production environment and click Save. 4. Call the number – your flow will run over the phone, with TTS + STT handled by Twilio. If you prefer to build your own voice interface (for example a mobile app or smart speaker prototype) you can stream audio through your own service and drive the conversation with the Dialog Manager API – that’s what some people refer to as “VAPI”. So the explanation you got from ChatGPT is only half-true: templates do start in chat mode, but voice channels are still available; you just need to hook them up explicitly. Docs & setup guides • Telephony/Twilio setup: https://docs.voiceflow.com/docs/setting-up-twilio-integration
• Voice limits & troubleshooting: https://docs.voiceflow.com/docs/troubleshooting-voice
• Dialog Manager API (custom interfaces): https://developer.voiceflow.com/reference/overview
Once you’ve connected Twilio you’ll get full TTS, ASR, barge-in, call recording (optional), and concurrency that matches your plan. Let me know if you don’t see the Telephony tab or need help with the Twilio webhook – happy to walk you through.
• New templates are chat-first, but you can turn any agent into a phone/voice bot by connecting the Telephony (Twilio) integration or by using the Dialog Manager API with your own audio front-end.
• No special beta or private access is required; you just need a Twilio account (a free trial works) and Editor rights in the workspace. What actually changed ───────────────────── The current builder uses a single “Agent” model. We ship most templates with a web-chat preview because that’s the fastest way to test. Voice is now treated as a deployment channel you add afterwards: 1. Open the agent ➜ Integrations ➜ Telephony.
2. Import or buy a Twilio phone number and save the SID/Auth Token.
3. Assign the number to your Development or Production environment and click Save. 4. Call the number – your flow will run over the phone, with TTS + STT handled by Twilio. If you prefer to build your own voice interface (for example a mobile app or smart speaker prototype) you can stream audio through your own service and drive the conversation with the Dialog Manager API – that’s what some people refer to as “VAPI”. So the explanation you got from ChatGPT is only half-true: templates do start in chat mode, but voice channels are still available; you just need to hook them up explicitly. Docs & setup guides • Telephony/Twilio setup: https://docs.voiceflow.com/docs/setting-up-twilio-integration
• Voice limits & troubleshooting: https://docs.voiceflow.com/docs/troubleshooting-voice
• Dialog Manager API (custom interfaces): https://developer.voiceflow.com/reference/overview
Once you’ve connected Twilio you’ll get full TTS, ASR, barge-in, call recording (optional), and concurrency that matches your plan. Let me know if you don’t see the Telephony tab or need help with the Twilio webhook – happy to walk you through.
eager-peachOP•3mo ago
I am probably building an unusual project- basically a sound board structure where I will pipe voiceflow audio (hopefully both AI text to speech [elevenlabs] and audio clips through a virtual cable on my PC so that I can switch between my live voice and voiceflow on sales calls. So, on the surface, it doesn't appear I need telephony. But, if it's necessary to build/test I guess I'd understand. Please advise.