Voice Model Testing

Side-by-side experimentation of native speech-to-speech models on the same scenario.

Transcript

--:--
first audio — turns 0 tools 0

Pick an arm and a scenario, connect, then talk.

Event trace

No source
1

Capture

Cost per turn

Where the money goes

At volume

Cost against latency

one point per session — n=1 each, so this is a spread, not a measurement