Inconsistent First Message Playback: Different results in dashboard vs. phone
Hello Vapi Support,
we are running into a major issue with the "First Message" audio playback:
Symptoms:
• The First Message plays perfectly in the dashboard, but is distorted or wrong-pitched on the phone (via Twilio) with the second use case it’s exact the other way arround.
• The behavior is not consistent: Sometimes it works in the dashboard and sometimes on the phone.
Test Cases:
ffmpeg -i input.mp3 -ar 44100 -ac 1 -c:a pcm_s16le output_vapi_recomendation.wav
• Vapi allows only one firstMessage file, which must work for both use-cases.
Key Question:
Which exact audio format (codec, sample rate, bit depth, mono/stereo) is 100% compatible for First Message, so it works cleanly both in the dashboard and in phone calls (Twilio)?
• Do you have an official “golden standard” or recommendation?
• Is it a known issue that PCM-mulaw (8kHz) fails in browser/dashboard?
• Can your backend deliver different formats depending on context (dashboard vs phone)?
• Or do you have a best-practice workflow for universal compatibility?
Thank you very much for your help!
we are running into a major issue with the "First Message" audio playback:
Symptoms:
• The First Message plays perfectly in the dashboard, but is distorted or wrong-pitched on the phone (via Twilio) with the second use case it’s exact the other way arround.
• The behavior is not consistent: Sometimes it works in the dashboard and sometimes on the phone.
Test Cases:
- output_vapi_smallest.wav
URL: https://cdn.jsdelivr.net/gh/bdm2m/first-messages/output_vapi_smallest.wav
o Format: WAV, mono, 8 kHz, µ-law (pcm_mulaw), very small file
o Result: Plays fine via phone/Twilio, but fails in the dashboard wrong pitched or slowed down - output_vapi_recomendation.wav
URL: https://cdn.jsdelivr.net/gh/bdm2m/first-messages/output_vapi_recomendation.wav
o Format: WAV, mono, 44.1 kHz (or 48 kHz), PCM Linear (pcm_s16le), typical web-optimized
o Result: Works in the dashboard, but fails in the via phone/Twilio wrong pitched or slowed down is distorted or silent via phone/Twilio.
Additional info:
• Source files are generated via ElevenLabs (MP3), then converted with ffmpeg, e.g.:
ffmpeg -i input.mp3 -ar 44100 -ac 1 -c:a pcm_s16le output_vapi_recomendation.wav
• Vapi allows only one firstMessage file, which must work for both use-cases.
Key Question:
Which exact audio format (codec, sample rate, bit depth, mono/stereo) is 100% compatible for First Message, so it works cleanly both in the dashboard and in phone calls (Twilio)?
• Do you have an official “golden standard” or recommendation?
• Is it a known issue that PCM-mulaw (8kHz) fails in browser/dashboard?
• Can your backend deliver different formats depending on context (dashboard vs phone)?
• Or do you have a best-practice workflow for universal compatibility?
Thank you very much for your help!
