Does Omnia handle realtime voice?

Yes. Duplex voice with interruption handling runs in the runtime, alongside a transcribe-think-speak pipeline when you want one.

Free while you build, the core free in production, and an enterprise licence per environment per year from £30,000. Nothing is metered per conversation or per line.

omnia vs cognigy

Cognigy and Omnia both run voice agents. Omnia also tests them as phone calls.

Omnia tests a release the way a caller hears it, then moves calls to it in steps without dropping the ones in progress. Cognigy’s simulator speaks text only, and restoring an earlier snapshot ends the calls already running.

A voice release goes wrong with two hundred calls live.

on omnia

The new version was answering a fraction of calls when its metrics dropped, and Omnia sent new calls back to the previous one. Callers already mid-conversation finished on the version they started with. docs

on cognigy

Restoring a snapshot overwrites the agent’s resources, and conversations in progress stop working because the flows are deleted. their docs

Simulate the phone call

Omnia’s test scenarios talk to the agent as audio and run on every release. A caller who cuts in halfway through a sentence is a scenario like any other, and each run is recorded for playback. docs

Cognigy: The Simulator only supports text-based conversations. source

Releases that don’t drop calls

A new version gets traffic in steps. When a pod is replaced, the sessions on it finish before it goes. docs

Cognigy: Deploying means selecting a snapshot on the endpoint. Traffic splitting isn’t documented, and a project keeps up to ten snapshots. source

Any Kubernetes, supported

Omnia installs with public Helm charts into the cluster you run, whichever distribution it is. docs

Cognigy: Self-install supports EKS and AKS; on-premises clusters such as OpenShift take significant extra configuration that Cognigy doesn’t support. source

At a glance

Topic	Omnia	Cognigy, per its docs
Simulated users	yes Multi-turn, against real tools. docs	partly LLM personas, text only. source
Adversarial simulated users	yes Yes. docs	not documented Not documented. source
Voice simulation	yes Barge-in, overlap, dead air, recorded. docs	not documented Not supported. source
Stepped rollout, auto rollback	yes Yes. docs	not documented Not documented. source
Tool policy	yes A CEL rule on every call’s arguments. docs enterprise	partly Per-tool conditions and MCP tool filters. source
End-user identity to tools	yes Carried to every call. docs	not documented Not documented. source
Agent traces over OpenTelemetry	yes Yes. docs	partly Voice Gateway call events only. source

Common questions

Does Omnia handle realtime voice?: Yes. Duplex voice with interruption handling runs in the runtime, alongside a transcribe-think-speak pipeline when you want one.
How is Omnia priced?: Free while you build, the core free in production, and an enterprise licencelicense per environment per year from £30,000. Nothing is metered per conversation or per line.

Put your own agent through the gate.

Install Omnia in your cluster and run it free while you build.

Cognigy claims are from NiCE’s published documentation, checked 22 October 2026. Spot something out of date? Tell us.

Install it in your cluster Or walk through it with us →