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 there, Thank you for your message. Our team is currently out of the office. We operate Monday through Friday, from 9:00 AM to 8:00 PM Pacific Standard Time (PST). We’ll get back to you as soon as possible during our normal business hours. If your message is urgent, please mark it accordingly or include “URGENT” in the subject line, and we’ll do our best to respond promptly. Warm regards, Vapi Customer Support Team

Hey @Jozef I’ve worked with real-time transcription and mixed audio streams before, and I can help you implement a solution that prevents your assistant’s TTS from being transcribed by ElevenLabs Scribe. To clarify the approach, can you confirm whether Vapi currently sends mixed mono/stereo audio to your custom transcriber, or does it keep separate channels for user vs. assistant?

Hi Jozef, Thanks for the detailed breakdown — here are clear answers to each of your questions based on how Vapi currently handles audio for custom transcribers: 1. Does Vapi send a mixed mono or stereo PCM stream? Vapi sends stereo, interleaved linear16 PCM audio to custom transcribers. 2. What is the exact channel mapping? The two audio channels are always mapped as: - Channel 0 (left): Customer audio - Channel 1 (right): Assistant TTS audio 3. Is there a way to receive only the user audio? At the moment, no — Vapi always combines customer + assistant audio into the stereo stream before sending it to the custom transcriber. 4. Is there any setting to prevent assistant TTS from being forwarded? There is no configuration that disables assistant audio from being included in the transcriber audio stream. Existing settings only affect how transcriptions are used, not what audio is sent. 5. Does Vapi support channel metadata for custom transcribers? Yes — custom transcribers can return a

channel

field (

customer

or

assistant

) in their transcription responses, and Vapi will respect it. However, Vapi does not send per-channel audio separately. --- Workaround for ElevenLabs Scribe v2 Because ElevenLabs Scribe v2 treats the stream as single-channel audio, the recommended approach is: Extract only the left channel (customer audio) in your transcriber server before sending audio to ElevenLabs.