Echo powers 60% cost savings on TTS
- Listens to your calls
- Chunks repetitive phrases
- Streams them from the cache instead of re-synthesizing them
Reuse the voice you've already paid your TTS provider for.
Early access is rolling out to teams building Voice AI with Pipecat.
01 — where the money goes
Voice makes the first impression. It also makes the bill go brrr…
Across a typical voice AI stack, text-to-speech is the single largest line item — bigger than your LLM and transcriber put together.
02 — how echo does it
Your agent says the same things all day long.
Greetings, confirmations, hold lines, tool calls, closings. Echo learns them once and you never pay for them again.
Listens
Echo sits inside your Pipecat pipeline and observes live calls — no rewrite, no new voice vendor.
Chunks & caches
It splits speech into reusable phrase-level audio chunks, keyed to the exact text your LLM produces.
Replays
Next time that phrase comes up, Echo rapidly streams the cached audio. No TTS call is made resulting in direct savings.
03 — one real call, annotated
Watch a call go by with Echo switched on.
Six agent responses. Only one of them ever reaches your TTS provider.
Call #4,181 — account activation
— call ends —
Echo's cache, filling up as the call runs.
Idle most of the call. Still your voice, still your quality.
04 — where it fits
Drops into Pipecat. Keeps whatever voice you chose.
Echo is a TTS Gateway for Voice AI agents and is TTS model agnostic. It caches the audio your TTS engine already produced, so nothing about your voice changes. Echo is now available on Pipecat.
Get early access
One email when we open your slot. Nothing else.