OpenRouter for Audio
One endpoint for every TTS, STT, and realtime voice model. Route by latency, provider, cost, or capability.
How it works
One Request. All Models.
One endpoint for every model. Allmodels normalizes provider schemas allowing you to use the latest frontier models with no need to integrate new APIs.
Why a router is faster.
A router sounds like an extra hop. For realtime voice, it's often the shortest path that matters.
Voice is a constant two-way stream of small audio frames. When your server is far from the user, every frame pays for that distance, along with jitter, stalls and retransmits. AllModels terminates and buffers the stream on a global edge close to the user, then opens clean upstream connections to your app and the model provider.
Without a router
Jitter, stalls and retransmits travel the whole way to a distant server.
With AllModels
The stream is buffered next to the user for about 30ms of routing overhead.