VoiceBench

Dashboard

Overview of your voice agents and test scenarios

0

AI-EndUserVoice

0

Saved Calls

0

Total Runs

Run ID	Calls	AI-EndUserVoice Name	Status	Overall Score	Initiated	Completed	Actions

Run Details

Call ID	Customer Number	Status	Scenario Name	AI-EndUserVoice Name	Call Duration	Overall Evaluation	Call Start Time	Language	Speed	Pitch Shift	Background Noise

Call Details

Response Latency from recording

Silence gap between user stopping and bot starting (measured from stereo WAV via local VAD)

Mean--

P90--

P95--

P99--

Samples--

Interruptions

AI interrupting user--

User interrupting AI--

Stop time after interruption--

Ext Agent Quality

Avg Pitch (Hz)--

Tone + Clarity (/5)--

Intent Verified (/5)--

Conversation

Talk Ratio--

Dead Air / Silence--

Hallucination #--

Call Info

Avg Ringing Duration--

Call Duration--

CSAT

Score--

Audio Quality

MOS--

UTMOS--

Recording

Transcript

0

Total Tests

0

Passed

0

Failed

—

Avg Score

Sessions

No sessions found. Run a test on the Testing page to see results here.

Performance Summary

Avg WER

—

Avg Accuracy

—

Avg Latency

—

Avg Quality

—

Create AI-EndUserVoice

AI-EndUserVoice Name

Persona & behaviour

AI Providers

TTS Provider

STT Provider

LLM Provider

Use LLM to shape spoken text

Parameters

Language

Cartesia Voice

Tone

Speech Speed 1.0x

Pitch Shift 0%

Acoustic Environment

Background Noise

Noise Level

Packet Loss None (0%)

Voice Modulation None

External Data Sources

Add URLs, CURL commands, or Python scripts. Results are stored in variables the agent can use during conversation. In the persona, write when to fetch — e.g. "When asked about balance, run {{fetch:account_balance}}".

List AI-EndUserVoice

AI-EndUserVoice ID	AI-EndUserVoice Name	Created	Last Executed	Last Edited	Language	Scenarios	Actions
No agents yet — create one from the sidebar.

AI-EndUserVoice

Persona & Behaviour

External Data Sources

🎧 Reference Conversation trains EndUser behaviour

Upload a real call recording or paste a transcript — VoiceBench will mirror this conversation style in every call.

Upload recording (WAV / MP3 / M4A)

Or paste transcript directly

Scenarios

AI-EndUserVoice Name	Language	Scenarios

Edit Scenario

Scenario Name

Caller Persona & Behaviour

Objective / Steps

Reference Transcript (optional — overrides agent-level reference for this scenario)

AI Providers

TTS Provider

STT Provider

LLM Provider

Use LLM to shape spoken text

Parameters

Language

Cartesia Voice

Tone

Speech Speed 1.0x

Pitch Shift 0%

Acoustic Environment

Background Noise

Noise Level

Packet Loss None (0%)

Voice Modulation None

Run Configuration

Total Runs

Execution Mode

Time Restriction (minutes)

Execution stops if time is exceeded; leave empty to run all

➕

Create New AI-EndUserVoice

Configure a voice agent for testing scenarios

AI-EndUserVoice Name

Role

Persona & Behavior

Voice

Tone

⚙️

Audio Settings

Configure real-world audio simulation

Environment

Noise Level

Packet Loss Medium (3%)

Voice Modulation Low

Default External AI-EndUserVoice

Default Language

Pipeline Configuration

Testing Console

Speech-to-Text Analysis

Text-to-Speech Analysis

Dashboard

Calls Dashboard

Run Details

Call Details

Response Latency from recording

Interruptions

Ext Agent Quality

Conversation

Call Info

CSAT

Audio Quality

Recording

Transcript ● Live

Sessions

Performance Summary

Create AI-EndUserVoice

AI Providers

Parameters

Acoustic Environment

External Data Sources

List AI-EndUserVoice

AI-EndUserVoice

Persona & Behaviour

External Data Sources

🎧 Reference Conversation trains EndUser behaviour

Scenarios

Scenarios

Scenario

Caller Persona & Behaviour

Objective / Steps

Edit Scenario

AI Providers

Parameters

Acoustic Environment

Run Configuration

Create New AI-EndUserVoice

Audio Settings

Transcript