WitAI

WitAI turns voice recordings into text using your own Wit.ai app. For each audio it returns what is said and the detected language.

When to use it

• Transcribe the voice notes customers send over messaging so an agent can answer them.
• Turn voicemail messages into text and store them in your CRM.
• Find out which language a customer speaks to route them to the right team.
• Let an agent transcribe audio when it receives it, using the component as a tool.

Configuration

This component needs a WitAi credential, with the access token and the server address of your Wit.ai app. Pick it in the component’s credential field. See how to create it in Credentials.

Field	What it does
Audio Input	The audio or audios you want to transcribe. Connect the file that reaches the flow, for example from a message with attachments. You can connect several.

Output

The Data output returns one record for each audio, with two fields:

• text: the transcribed text.
• language: the language detected in that text.

Connect it to an agent, a text template or any component that works with text.

If the audio is not recognised

The component reports an error when it can’t understand the audio. This usually happens if the language of the recording doesn’t match the one set in your Wit.ai app, or if the audio is very noisy or unintelligible.

Example

A customer leaves a voice note asking to change their delivery date. Your flow passes the audio to WitAI, which returns the text “I want to move the delivery to Thursday” and the language English. An agent reads the text, finds the order and replies to the customer.

Related components

• OpenAI Whisper: another way to turn audio into text.
• To listen to the caller during a call, use the Voice Agent’s Wit.ai STT node.