WhatsApp voice notes
Let customers send voice notes and let your WhatsApp agent answer with one. Choose when it speaks, how it hears, which voice it uses and what happens if voice fails.
What you will achieve
Your agent will understand voice notes that customers send, and, if you wish, answer with a voice note of its own.
Prerequisites
- A WhatsApp agent.
- For spoken replies, a voice provider available on your server. If it is not, the Voice field says ElevenLabs is not available right now.
Choose how it replies
Open the agent and find the Voice notes panel on its Assistant tab.
| Replies | What the agent sends |
|---|---|
| Always text | Text only. |
| Voice when the customer sends voice | A voice note when the customer's message was a voice note, text otherwise. |
| Always voice | A voice note whenever it can. |
| Let the agent decide | The agent chooses, case by case. |
Choose a Voice from the list. Without a voice chosen, the agent sends text.
Choose how it hears
Hearing customers' voice notes picks the speech-to-text service: GPT-4o Mini Transcribe, Deepgram Nova 3 (English) or Deepgram Nova 3 Multilingual. Pick the multilingual one if your customers speak several languages.
The inbox shows each voice note with its written transcript, so you can read what was said.
When voice is not used
The agent falls back to text, and the customer still gets an answer, when:
- the reply is long (over about 900 characters; spoken replies are kept short);
- no Voice is chosen for the agent;
- the voice service is not set up, fails or does not answer;
- the customer's voice note is longer than five minutes or cannot be understood.
Costs
Hearing and speaking are charged from your credit with the reply, based on the length of the note and of the spoken text. They appear in Analytics and costs under the cost of the conversation.
Something here does not match what you see in the dashboard? Let us know.