Voices
18 voices. Every accent that calls.
Vexa simulates callers across General American, British, Australian, Irish, Scottish, Nigerian, Kenyan, Indian, French, German, Mandarin-accented English, and US Spanish bilingual. If real callers have an accent, your agent needs to be tested on it.
Most voice agent teams test on one or two voices. Real caller populations span dozens of accents with meaningfully different ASR error rates, prosodic patterns, and barge-in timing. Vexa exposes those differences before real callers do.
Every voice below is available in the default voice pool. Pass a subset to the config.voices array in a run to test against a specific group. The API reference documents the GET /voices endpoint if you want to pull the list programmatically.
Scale plan: bring your own recorded voice or a custom TTS voice to simulate a specific caller segment. Contact sales@vexavoice.tech to discuss custom voice integration.
North American English
Five voices covering the major US dialect regions: from General American to New England, Southern, Midwestern, and Texan.
Alex
alex
General American
en-USCard replacement flow
Male-presentingBaseline voice for most suites. Neutral, unregional accent.
Eleanor
eleanor
New England
en-USAccount inquiry
Female-presentingNon-rhotic vowels and clipped cadence. Tests r-dropping behavior in ASR.
Marcus
marcus
Southern US
en-USClaims escalation
Male-presentingSlower tempo, drawn vowels. Common failure point for WER on banking agents.
Quinn
quinn
Midwestern US
en-USTransfer request
Gender-neutralFlat intonation, full rhoticity. Useful as a stability baseline next to regional voices.
Sadie
sadie
Texan English
en-USPassword reset
Female-presentingDiphthong shifts and deliberate pacing. Tests fast-barge-in recovery.
British Isles
Three voices spanning standard British, Scottish, and Irish English. Non-rhotic accents with distinct vowel systems.
Priya
priya
British English (SE)
en-GBBooking confirmation
Female-presentingStandard Southern English RP. Strong non-rhotic cues expose ASR gaps.
Callum
callum
Scottish English
en-GBRefund dispute
Male-presentingRhotic with distinct vowel length contrasts. A consistent high-WER trigger.
Niamh
niamh
Irish English
en-IEEmergency transfer
Female-presentingRhotic with melodic intonation. Tests how agents handle question intonation on statements.
Australia and New Zealand
Two voices from the Southern Hemisphere Anglophone region, with distinct vowel raising patterns.
Dev
dev
Australian English
en-AUOutage alert
Male-presentingRaised TRAP vowel and high-rising terminal. Tests question detection in TTS output.
Aroha
aroha
New Zealand English
en-NZSubscription cancel
Female-presentingKIT vowel fronting and DRESS raising. Distinct from Australian despite geographic proximity.
African Anglophone
Two West and East African English voices with distinct rhythmic and tonal patterns.
Chidi
chidi
Nigerian English
en-NGLoan inquiry
Male-presentingSyllable-timed rhythm and distinct vowel quality. Exposes bias in US-trained ASR models.
Fatima
fatima
Kenyan English
en-KESIM swap request
Female-presentingSwahili phonological influence on stops and vowels. Tests multi-lingual ASR configurations.
South Asian
One South Asian English voice reflecting the world's largest English-speaking population. High WER impact on US-optimized models.
Ananya
ananya
Indian English
en-INFlight rebooking
Female-presentingRetroflex consonants and syllable timing. One of the highest-WER voices on standard ASR models.
Rohan
rohan
Indian English (North)
en-INInsurance claim
Male-presentingHindi phonological substrate with aspirated stops. Tests agent behavior under high-WER conditions.
European-accented English
Two European-accented English voices representing French and German L1 speakers. Common in enterprise support scenarios.
Camille
camille
French-accented English
en-FRWarranty claim
Female-presentingFront vowel shifts and consonant substitutions from French phonology. Tests ASR on L2 speakers.
Matthias
matthias
German-accented English
en-DEEnterprise support
Male-presentingPrecise consonants and flat prosody. Tests tone detection on voices with low prosodic variation.
East Asian-accented English
One Mandarin-accented English voice representing a high-volume L2 caller segment in technology and banking.
Wei
wei
Mandarin-accented English
en-ZHBilling inquiry
Male-presentingTonal prosody carry-over and fricative substitutions. High-WER trigger on financial vocabulary.
US Spanish Bilingual
Two US Spanish-English bilingual voices for code-switching scenarios. Critical for any US-facing voice agent.
Lucia
lucia
US Spanish bilingual
en-ESInsurance claim (code-switching)
Female-presentingSwitches between English and Spanish within a single call. Tests multilingual ASR and agent routing.
Rafael
rafael
US Spanish bilingual (Miami)
en-ESAppointment booking
Male-presentingMiami Spanish substrate with distinct vowel inventory. Second bilingual voice for suite diversity.
Custom voices on Scale
The 18 voices above are available on every plan. Scale customers can bring additional voices:
- Recorded callers. Upload a set of real caller recordings and Vexa builds a synthetic voice that preserves the accent, pacing, and vocal texture. Useful when you have a known high-volume caller segment.
- Custom TTS voices. Bring a voice from ElevenLabs, Play.ht, or a provider you already license. Vexa wraps it in the simulation pipeline and scores it the same way as built-in voices.
- Internal personas. Model specific caller types from your own user research: age cohort, cognitive accessibility pattern, non-native-speaker profile.
Custom voices are scoped to your workspace and not shared across customers. To discuss custom voice integration, contact sales or email sales@vexavoice.tech.
Use voices in a run
Pass a config.voices array to POST /runs to run a suite against a subset of voices. If omitted, Vexa uses all voices assigned to the suite.
curl -X POST https://api.vexavoice.tech/v1/runs \
-H "Authorization: Bearer vexa_sk_live_xxxx" \
-H "Content-Type: application/json" \
-d '{
"agentId": "agent_01J9KXPR7WQVHB2N5M8ZD3CA4T",
"suiteId": "suite_card_replacement",
"config": {
"voices": ["chidi", "ananya", "lucia", "wei"]
}
}'To list all available voice IDs with their metadata:
curl https://api.vexavoice.tech/v1/voices \
-H "Authorization: Bearer vexa_sk_live_xxxx"
# Filter by region
curl "https://api.vexavoice.tech/v1/voices?region=british-isles" \
-H "Authorization: Bearer vexa_sk_live_xxxx"See the full voices endpoint documentation for response shape and filter options.
Every voice, tested.
Run your agent through all 18 voices and see where task success drops. Free to start.