00:00How it works

A call, second by second

Here is what Rescuora does during one call, who is responsible at each step, and what your center configures before the first shift.

Call 4127 · timingLiveExample
00:00
Pickup
00:01
First caption
00:03
Language detected, translation on
00:20
CAD draft ready to confirm

ExampleDesign targets for one call; pilots will measure them.

00:04The steps

Six steps, each with a time budget

  1. 00:00

    Audio forked

    Your phone system sends a copy of the call audio to the appliance. The call itself is untouched.

    Target: no added delay to the call
  2. 00:01

    Caption

    Lingora captions the caller and call-taker in separate lanes.

    Target: under 1 s behind the voice
  3. 00:03

    Language and translation

    If the caller is not speaking the center's language, translation starts. Read-back speech is ready for the call-taker to trigger.

    Target: language detected in 3 s
  4. 00:03

    Sounds

    Sonara tags sound events and rising distress with a word and a time.

    Target: tag within 2 s of the sound
  5. 00:05

    Next question

    Protora suggests the next question from your protocol. The call-taker asks it, rewords it or skips it.

    Always a suggestion
  6. 00:20

    CAD draft

    Fields fill from the conversation. Each one is sent to CAD only after the call-taker confirms it.

    Always confirmed by a person

00:09Who does what

Rescuora suggests; people decide

Responsibilities in the first release
TaskRescuoraCall-takerSupervisor
Talk to the callerNoYesCan join
TranslateCaptions and drafts read-back speechTriggers read-back, can bridge an interpreterReviews later
Recognise soundsTags and times themDecides what they meanReviews tags in QA
Choose the next questionSuggests from your protocolAsks, rewords or skipsSets which protocols load
Fill the CAD recordDrafts fieldsConfirms each fieldAudits changes
Send unitsNeverThrough CAD, as todayAs today

00:15Configuration

What your center sets up before the first shift

protocols.yaml · excerpt
protocol_source: center-licensed  # your protocol set, loaded by your staff
suggest:
  mode: next_question
  max_on_screen: 1
cad_draft:
  fields: [location, incident_type, callback, hazards]
  require_confirmation: true  # cannot be turned off
read_back_tts: call_taker_trigger

ExampleIllustrative configuration; field names will change.

What we need from you
  • Your protocol set, in the format your vendor licenses to you
  • A media fork from your phone system (SIP/RTP)
  • A CAD test environment and its interface specification
  • The languages your callers speak most, from your own records
  • Two supervisors to review shadow-mode calls each week

00:22Outcomes we will measure

What a pilot is designed to show

We will not publish results until pilot centers have reviewed them and agreed to share them.

Pilot measure

Time

From pickup to a confirmed location, with and without Rescuora.

Pilot measure

Waits

Calls that needed an interpreter line, and how long the caller waited.

Pilot measure

Keystrokes

Typing per CAD record, and fields corrected after confirmation.

00:30Questions

How-it-works questions

A person on our product or engineering team answers every message.

Q01What if a caption is wrong?

The call-taker hears the caller too, and the original words always sit above the translation. Corrections in review feed the next model update.

Q02Can a suggestion be sent to CAD without a person?

No. Confirmation by the call-taker is required for every CAD field, and the setting cannot be turned off.

Q03Does it work on text-to-911?

Planned. Translation of text sessions is on the roadmap after voice.

Dispatch

Watch it run on a sample call

The console on our home page plays a scripted call in three languages. Then talk to us about your floor.