voice AI platforms like Vapi commonly use a two-channel audio setup for a clear and practical reason: to separate the speakers in a conversation. In a typical Vapi call, one channel is dedicated to the user's audio (input), and the second channel is for the AI agent's audio (output). so vapi uses 2 channels to deepgram on eeach request like you can see here: /v1/listen?punctuate=true&numbers=true&no_delay=true&encoding=linear16&channels=2&multichannel=true&filler_words=true&interim_results=true&language=en&model=nova-3&sample_rate=16000&mip_opt_out=false&endpointing=300&numerals=false Request details Endpoint /listen Deployment Hosted Status 101 OK Method Streaming Encoding: linear16 Channels: 2 Sample Rate: 16000 Performed at 06/11/2025 21:23:29 Summary Audio duration 00:02:13 Tier nova-3 Cost $0.03424 Inputs Models 40bd3654-e622-47c4-a111-63a61b23bfe8 Tags - Features Endpointing True filler_words True Interim Results True Multichannel True Punctuate True -- the Channels: 2 i only want 1 channel, i dont want a separate channel for usser and agent audio, can i change this to use 1 channel? because my cost gets double with deep gram if i use 2 channels i am using external api keys for openai and deepgram. you can see on this call: b4c762a1-bb9c-4b28-a58f-9817f4150f11

Currently, all official Vapi documentation and code examples use a 2-channel setup for Deepgram (one for user, one for agent), and there is no explicit guidance on switching to a single channel. To help you further: Are you open to using a custom transcriber implementation, or do you specifically need a solution within the standard Vapi configuration? Source: - [Vapi Custom Transcriber Guide](https://docs.vapi.ai/customization/custom-transcriber)

Hey Gurinder Singh, by default, we have enabled it for 2 channels. You have to raise a feature request for enabling it for a single channel only. You can raise an feature request for it over here: https://roadmap.vapi.ai