Concepts
VOIX converts a single call recording into a complete, structured record of the conversation. This section describes each component of that record and how to interpret it.
Call processing sequence
A call arrives
A telephony or call-recording system submits the call to VOIX, first when the call begins and again with the audio once it ends. See the Ingest reference for the submission process.
The call is transcribed
The recording is converted to text. Each speaker is separated onto individual turns, and the spoken language is detected automatically. See Transcripts.
Tone is measured
As the transcript is produced, VOIX assesses the sound of the conversation: the energy, control, and positivity in each speaker’s voice over time. See Tone signals.
The call is analyzed
VOIX reads the full conversation and produces a title and summary, a set of scores, answers to any custom checks, and a list of follow-up tasks. See Analysis & scores.
Participants are identified
VOIX matches the outside party to a contact and, where possible, attributes the call to the team member who handled it.
Output of a completed call
Every completed call produces the following:
- A speaker-separated transcript in the language of the call.
- A short title and summary describing the substance of the call.
- Six performance scores and a success flag, each with a written explanation.
- A tone reading per speaker, including how the mood shifted from the first half of the call to the second.
- Answers to any configured custom checks, each with a supporting quotation.
- Follow-up tasks derived from commitments made during the call.
- The contact on the other end of the call and, where clear, the team member who handled it.