Detect every glitch, pause,
and hallucination in your voice AI
Upload audio, train custom models, get timestamped detections. Built for teams shipping production voice experiences.
Trusted by Voice AI teams
ElevenLabs
PlayHT
Resemble
Deepgram
AssemblyAI
WellSaid
Murf
SpeechifyElevenLabsPlayHTResembleDeepgramAssemblyAIWellSaidMurfSpeechifyHow it works
Three steps from raw audio to production-grade artifact detection.
Upload labeled audio
Upload labeled audio
Drag in your audio files with artifact annotations. WAV, MP3, FLAC — we handle the encoding.
Train a custom model
Train a custom model
Our ML pipeline encodes, trains, and validates automatically. Your model is ready in under 2 minutes.
Get timestamped detections
Get timestamped detections
Run inference on new audio. Every artifact is pinpointed with start/end times and confidence scores.
Everything you need to ship with confidence
Purpose-built tools for voice AI quality assurance, from annotation to detection.
Custom Models
Train artifact classifiers on your own labeled data. Each model learns your specific voice AI's failure modes.
Timestamped Detections
Get precise start/end timestamps for every detected artifact — down to 50ms granularity.
API-First
Full REST API with presigned uploads, webhook notifications, and comprehensive SDKs.
Multi-Artifact Detection
Detect glitches, unnatural pauses, TTS hallucinations, clipping, and more — all in a single pass.
Fast Training
Models train in under 2 minutes on GPU-accelerated infrastructure. Iterate quickly on your detection pipeline.
Usage-Based Pricing
Pay for what you use. Generous free tier, transparent per-minute pricing, no hidden fees.
API-first by design
A simple REST API with presigned uploads and structured responses. Ship an integration in an afternoon.
curl -X POST https://api.relayai.dev/v1/inference-jobs \
-H "Authorization: Bearer rl_sk_..." \
-H "Content-Type: application/json" \
-d '{
"model_id": "mod_abc123",
"config": {
"threshold": 0.7,
"min_duration_ms": 100
}
}'{
"id": "inf_xyz789",
"status": "completed",
"files": [{
"filename": "call_recording.wav",
"detections": [
{
"type": "glitch",
"start_ms": 3200,
"end_ms": 3450,
"confidence": 0.94
},
{
"type": "unnatural_pause",
"start_ms": 8100,
"end_ms": 8900,
"confidence": 0.87
}
]
}]
}Simple, transparent pricing
Start free, scale as you grow. No hidden fees, no surprises.
Starter
For teams getting started with voice AI QA
- 500 minutes of inference/month
- 10 training jobs/month
- 5 custom models
- $0.12/min overage
- API access
- Email support
Pro
For production voice AI teams
- 2,500 minutes of inference/month
- 50 training jobs/month
- Unlimited models
- $0.10/min overage
- Priority API access
- Slack support
- Webhook notifications
- Team management
Enterprise
For high-volume and custom requirements
- Unlimited inference
- Unlimited models
- Unlimited datasets
- Dedicated infrastructure
- SLA guarantee
- SSO / SAML
- Custom data retention
- Dedicated support engineer
Start detecting artifacts today
Free trial, no credit card required. Get your first detections in under 5 minutes.