Audio Translation
Audio Translation
The Audio Translation service translates spoken language on a call in real time, so two people who speak different languages can talk through SBN.
Overview
While a call is up, Audio Translation listens to the speech, recognises the language, and translates it - one-way or both ways - producing the translated words as they are spoken. It can be used on a call set up within SBN, or on a dedicated dial-in line that a caller rings to get translation.
The speech recognition and translation are done by a cloud provider (Soniox). When a translation call starts, a short spoken intro message can be played first.
- Runs: When configured. One per host.
- Required: Only where live spoken-language translation is needed. Not needed otherwise.
- Depends on: NATS; the Soniox service (with an API key and outbound internet access); the Text to Speech service for the spoken intro; and the SIP services for dial-in lines.
Configuration
Audio Translation is configured under the audioxlate namespace. View the defaults with ./sbn-media config eject and set overrides in sbn-media.local.yaml.
| Setting | Default | Description |
|---|---|---|
audioxlate.apikey | (none) | The Soniox API key. Required. |
audioxlate.model | stt-rt-v3 | The Soniox model used for recognition and translation. |
audioxlate.languages | [no, sv, da, en] | The languages to expect, used to help recognise which is being spoken. |
audioxlate.defaultLanguage | en | The language assumed when a line does not specify one. |
audioxlate.realm | (none) | The SIP realms whose calls may request translation. |
audioxlate.dialInLines | (none) | SIP lines a caller can ring to get translation directly. |
audioxlate.introDelayInMs | 250 | How long to wait before playing the intro message. |
audioxlate.messages | (per language) | The intro and error messages, per language, spoken to the caller. |
After changing these in sbn-media.local.yaml, apply them with ./sbn-media config reload.
FAQ
What do I need before it will work?
A Soniox API key in audioxlate.apikey, outbound internet access to Soniox, and - for the spoken intro - the Text to Speech service configured. For dial-in use, set up the lines in audioxlate.dialInLines.
Which languages can it handle?
It is told which languages to expect in audioxlate.languages to help it recognise them; the provider does the recognition and translation. Set defaultLanguage for when a line does not specify one.
How does a caller reach translation directly?
Through a dial-in line listed in audioxlate.dialInLines - the caller rings that line and translation starts, beginning with the intro message.
Can the spoken messages be changed?
Yes - the intro and error messages are set per language in audioxlate.messages.
Related pages
- SBN Media Overview (
SBN-Media/overview) - Installing and Configuring SBN Media (
SBN-Media/installation) - Text to Speech (
SBN-Media/Media-Processing/text-to-speech) - SIP Server (
SBN-Media/Telephony/sip-server)