Act on speech as it happens.

Give your voice agent the context it needs, while the conversation is still unfolding.

We build Accel1, our own speech model. It turns live audio into actionable signals developers subscribe to, so their agents can understand more, prepare earlier, and spend fewer reasoning turns working out what the audio already reveals.

Hear more than words.

We train Accel1 to recognize intent, timing, and speaker cues in live speech. Choose the signals your application needs, or define your own, and feed them into your agent as the conversation unfolds.

Live audio

The conversation
as it unfolds

Accel1

Our speech model

Signals

Turn-taking
Intent & urgency
Speaker cues
Security & custom signals

Your agent

Any model.
Any pipeline.

Earlier context

Prepare or respond while the caller is still speaking.

Richer understanding

Bring timing and speaker cues into your agent’s context.

Programmable behavior

Define what to listen for and how your agent responds.

Available in our cloud or on premises, within your own infrastructure.

Discuss deployment

Signals that move the conversation forward.

Natural turn-taking

Recognize when the caller is finished, still thinking, or interrupting, so your agent can speak at the right moment.

Intent and urgency

Surface what the caller needs while they are speaking, giving the agent context to prepare, route, or escalate.

Call connection and hold monitoring

Signal a live answer, a recording, or a representative returning after hold music, so the conversation can move forward.

Security during the call

Subscribe to multiple signals for defined security concerns as they surface, and route them into your application’s review or escalation flow.

Robotics and voice-controlled work

Surface spoken instructions, pauses, and corrections while a task is in progress, giving the controller timely context.

Smart home interactions

Recognize requests and changes of mind about home devices as they are spoken, so the application can adapt.

The team

Michael Liberman

Co-founder

Michael is an AI and technology executive with a track record of building and leading AI teams, shaping technology strategy, and turning complex ideas into successful products. He previously led the Vonage AI team (part of Ericsson), developing conversational AI technologies and advanced audio models. His experience bridges deep technology, product vision, engineering leadership, and business execution.

Michael Patkin

Co-founder

Michael brings over 15 years of experience in applied machine learning, AI research, and engineering. He combines deep expertise in model architecture, training and fine-tuning, and inference optimization across audio and language domains with a strong track record of leading technical teams. He excels at translating advanced machine learning research into efficient, scalable, production-ready AI systems.

What should your agent hear?

Tell us what you are building and which signals from a live conversation would help your agent respond.