tazo/labs
Contact salesAPI Console

Speech to Decision
Skip the Transcript

Tazo is an audio decision model It hears the conversation and decides what happens next, without waiting for text to arrive

Try Tazo

Trusted by developers from

FAQs

Something we missed? Write to our team.

What is an audio decision model?
An audio decision model listens to audio and picks the next action from a set of options, with a calibrated probability for each. It doesn’t generate text. In a voice agent, it decides when to respond, wait, backchannel, yield or escalate. Tazo is an audio decision model built for live calls.
How is Tazo different from an LLM?
An LLM generates text one token at a time, and in a voice agent it only runs once the turn is over. Tazo picks from options you define, straight from the audio, and returns a probability for each. Your LLM decides what to say. Tazo decides when, and what happens next.
Does Tazo need a transcript?
No. Tazo reads raw call audio, so it hears the pitch, pauses and tone that a transcript drops. Those cues are often what tells you whether a caller has finished or is still thinking. Why the voice AI stack needs a decision model.
Which decisions can Tazo make?
Any choice you can list: respond or wait, keep talking or yield, backchannel, voicemail, escalate to a human, or start a tool call early. You define the options in the request, so a new decision is a new option list, not a new model. See the decisions a call needs.
How does Tazo fit into my voice AI stack?
Tazo runs as a standalone API next to your existing STT, LLM and TTS. It reads the same call audio and sends back decisions, so you keep your models, prompts and tools.
What are calibrated probabilities, and why do they matter?
A calibrated probability means confidence matches reality: decisions made at 0.9 should be right about nine times in ten. That lets you set a threshold per action, for example escalating to a human only above 0.9.
How is Tazo different from Jev or AudioJev?
Jev decides from text, so a voice agent has to transcribe the call first. AudioJev answers questions about short audio clips, and its scores aren’t calibrated. Tazo is built for live calls: audio in, calibrated decisions out. Read the full comparison.

Get API access

Contact sales

Tell us a bit about what you’re building and we’ll get back to you.

Prefer email? support@tazolabs.com