Audelyr product

Preview

A runtime for responsive voice agents

Coordinate listening, reasoning, tools and speech with interruption-aware session events.

$0.060/minillustrative usage rate
<300 mstarget first result
61+illustrative languages

Capabilities

One interface.
Clear boundaries.

Designed as composable infrastructure, with Preview labels wherever production behavior is not yet available.

01

Turn detection

Documented, observable and designed to fit into your existing product architecture.

02

Barge-in events

Documented, observable and designed to fit into your existing product architecture.

03

Tool calls

Documented, observable and designed to fit into your existing product architecture.

04

Session state

Documented, observable and designed to fit into your existing product architecture.

05

Observability

Documented, observable and designed to fit into your existing product architecture.

06

SIP roadmap

Documented, observable and designed to fit into your existing product architecture.

Interface preview

Start with a sample.
Inspect every field.

The Preview returns simulated data and never implies that customer audio was processed.

const job = await audelyr.preview.transcriptions.create({
  sample_id: "multilingual-support-call"
});
console.log(job.preview, job.transcript);
01

Built for live systems

A streaming-first interface keeps application state, timing and model output in one observable session.

  • Stable event types
  • Regional processing controls
  • Usage and latency metadata
02

Formats and limits

Preview interfaces document PCM, WAV, MP3 and Opus workflows. Published limits are illustrative until production inference launches.

  • 16–48 kHz audio
  • Mono and stereo input
  • WebSocket and asynchronous patterns
03

Security by design

Keys are hashed at rest, access is scoped per account, and customer audio is not accepted by the current Preview API.

04

Frequently asked questions

The public Preview demonstrates interface shape and developer workflow. It does not perform real speech inference and is not a contractual service offer.

Early Access

Build the interface
before the infrastructure.

Explore the honest Preview, then tell us what production voice needs from Audelyr.

A runtime for responsive voice agents — Audelyr