Your concierge takes a booking in 45 languages.
Guests call from a loud lobby, in Gulf Arabic, Brazilian Portuguese, or South Indian English — and they want a king suite, late checkout, and to feel welcome. Roark scores every call on the audio: accent clarity, warmth, the booking detail, the upsell.
Backed by YCCaller: I want a room for the thirtieth of Ramadan.
Agent: I'm sorry, did you say the thirteenth?
Accent breakdown — Gulf Arabic caller misheard — heard by the audio model, invisible to the transcript.Accent clarity
Scoring production voice AI for teams at
§01 · When the call goes wrong
Here's how a guest call goes wrong.
Each one is a botched booking, a guest who felt unwelcome, or revenue left on the table — and most are inaudible to a tool that only reads the transcript.
The accent it could not parse
Caller: I want a room for the thirtieth of Ramadan.
Agent: I'm sorry, did you say the thirteenth?
A Gulf Arabic or South Indian English caller is misheard, the date flips, and the booking is wrong from the first turn. Accent breakdown is an audio failure a transcript-only tool reads as a clean exchange.
Accent clarityThe wrong room, confidently confirmed
Caller: Two queens, non-smoking, two nights.
Agent: Confirmed — one king, smoking, for two nights.
The agent read back the wrong details with total confidence. Roark checks the confirmed booking against what the guest actually asked for, every call.
Booking accuracyWarm request, robotic reply
Caller: It's our anniversary — can you do anything special?
Agent: Anniversary noted. Is there anything else.
The words were fine; the delivery was cold and flat on the one call that should have felt personal. The audio model scores warmth and emotion the text can never show.
WarmthThe suite it never offered
Caller: Do you have anything with a view?
Agent: Yes, our standard rooms face the courtyard.
A clear buying signal, and the agent never offered the ocean-view suite or late checkout. Roark scores whether the agent surfaced the upgrade the guest was reaching for.
Upsell captureDrowned out by the lobby
Caller: ...checking in under Okafor, party of four...
Agent: Sorry, could you repeat the name three more times?
Background noise from a packed lobby buried the guest, and the agent collapsed into repeat-loops instead of recovering. Roark scores noise robustness and ASR accuracy on the real audio, not the cleaned-up transcript.
Noise robustness§02 · From caught to fixed
Roark catches every one of these — and proves the fix.
Each failure above is filed with its evidence, becomes a repeatable simulation until a candidate passes, and is verified on your next thousand live calls.
Your fix, replayed against the exact failures above.
Every change explicit and diffed — you apply it.
You ship — Roark confirms the metric moved on live calls.
you ship it — Roark verifies every call, in 45 languages
…and the loop runs again on the next call.
§03 · Simulate before launch
Break it in staging,
not in production.
Run your agent against hundreds of simulated callers — realistic personas, accents, background noise and edge cases — and get every conversation scored before a customer ever dials in.
Scenarios & personas
Hundreds of simulated callers — the angry one, the rambler, the interrupter — built from your real call types.
45 languages & accents
Native accents, code-switching and background noise — in every market your agent answers.
Load & health tests
Peak-volume concurrency and always-on health checks, so the agent that passed in staging survives launch day.
Run it in CI
Every prompt or model change runs the suite before it merges — quality gates for conversations, not just code.
1 failure filed as an issue — fix it before launch, not after
§04 · Evals & observability
64+ metrics. Your models,
not just an LLM.
Every production call scored as it lands — issues filed, alerts fired, dashboards and OTEL traces on tap, for voice calls and chat threads alike. And where most tools grade a transcript with an LLM, Roark runs purpose-built audio models on the call itself, measuring what your customer actually heard.
Everyone else
LLM reads the transcript
“The agent said the right words.” Misses how it sounded — the mispronounced drug name, the flat apology, the rushed close.
Audio models hear the call
Pronunciation, accent, emotion and vocal stress measured from the waveform — the signal an LLM grading text can never see.
Built for 45 languages
The way the world actually calls.
Native accents, code-switching and local noise — not English with a filter. Every audio-native metric runs in each one.
growing every release
Audio-native
custom models
- Accent clarity
- Warmth
- Emotion
- Noise robustness
- Pronunciation
- Pace & pauses
Languages
45 languages
- Multilingual ASR
- Language match
- Dialect handling
- Code-switching
Conversational
LLM + rules
- Booking accuracy
- Upsell capture
- Task success
- Hallucination
- Repetition
- Tone
Performance
latency
- Time-to-first-word
- Turn latency
- ASR WER
- Barge-in handling
§05 · Get started
First call scored in under a minute.
One click on any platform below and production calls stream in on their own — or send any recording with three lines of code.
Read the quickstartimport Roark from '@roarkhq/sdk'const roark = new Roark({ apiKey })await roark.calls.evaluate({recordingUrl, agent: 'support_v2',}) // scored in seconds
Works with
Also built for
Customer Support
Resolve it for real, escalate cleanly, and hear the frustration the transcript hides.
ExploreRetail
Quote the policy as written, never invent stock, and hold your brand voice when a returns caller is furious.
ExploreHealthcare
Say the drug name right, verify before PHI, and never sound like a robot to a scared patient.
Explore
HIPAABAA availableGuest data & payment handling
Roark scores PCI-DSS card-handling scripts as pass/fail, redacts captured card numbers, and runs with configurable retention for guest PII.
Bring a recording.
We’ll score it live.
See your own agent measured on the audio it actually produced — in the demo, in real time. Stop guessing whether your voice AI works.
founders@roark.ai · we reply fast