Speko

Route voice models across multiple spoken languages

Pricing model

Demo

Rating

★ People love it

Time to value

Quick Setup (< 1 hour)

You can use Speko to route speech-to-text, text-to-speech, and language model requests to the best performing model for each language. Instead of relying on general English benchmarks, the service tests dozens of voice models across multiple languages to measure accuracy and response times. Developers can connect their voice agents through a single API or use the open gateway runtime for framework integrations like LiveKit and Pipecat. This allows you to select the fastest and most accurate provider for every user conversation without writing custom switching logic.

What Speko does

Route voice sessions to top performing models based on languageConnect voice agents using LiveKit or Pipecat librariesCompare word error rates across speech-to-text providersMeasure model response times and latency across languagesConfigure bring-your-own-key credentials for model providersConnect development environments to voice benchmarks via MCPLanguage-by-language benchmarking across dozens of speech modelsAutomatic routing to the most accurate speech-to-text model per languageHosted API data plane for speech-to-text, LLM, and text-to-speech callsOpen customer-side runtime gateway for direct provider streamingBring-your-own-key credential support for external model providersNative integrations with LiveKit and Pipecat voice frameworksModel Context Protocol integration for Claude and Cursor