
1 / 3
【Overview】 Speko is a voice AI router designed for developers building voice agents. It benchmarks 79 speech and language models across 17 languages and routes each voice session to the model that performs best for that language and objective. The router is exposed through one hosted API covering speech-to-text, LLM, and text-to-speech. The problem Speko solves is model selection: instead of manually testing and hardcoding providers, teams get measured routing that picks the winning model per session. This matters for teams whose users speak languages other than English, where a single model may not deliver the best results across the board. Speko is aimed at developers building voice agents and teams that need multilingual voice support. It provides a unified API, framework integrations for LiveKit and Pipecat, and a customer-side runtime with provider-direct streaming and local bring-your-own-key (BYOK) credentials. 【Key features】 • Measured routing: Benchmarks 79 speech and language models across 17 languages and routes each session to the model that wins for that language and objective. • Router STT: Transcribes an uploaded audio file or opens a realtime transcription WebSocket, returning the concrete provider and model used. • Router LLM: Generates typed JSON or SSE LLM responses with managed selection or an explicit provider and model. • Router TTS: Synthesizes one utterance to a raw audio stream or opens a streaming synthesis WebSocket. • Gateway runtime: Provides an open customer-side runtime with provider-direct streaming, local BYOK credentials and optional Speko-managed routes. • Framework integrations: Ships native LiveKit and Pipecat Python integrations plus a Claude/Cursor MCP bridge. • Pre-response failover: Fails over to another provider before any output is returned, exposing the serving attempt via Speko-Attempt-ID. • Phone calls: Places phone calls on Speko numbers with carrier charges passed through. 【Use cases】 1. Route voice sessions to the best model per language 2. Transcribe audio with managed provider selection 3. Generate LLM responses for voice agents 4. Synthesize speech 5. Build and observe LiveKit and Pipecat voice workers 【Pricing】 Router · +5% on the provider's rate — Speech-to-text, LLM, TTS at the provider's published rate, Measured selection and pre-response failover, Public preview access, Email support Speko infra · $0.09/minute — Speech-to-text, LLM, TTS included, Measured selection and pre-response failover, Public preview access, Email support Enterprise · Custom locked for the term — Speech-to-text, LLM, TTS included, Measured selection and pre-response failover plus deployment reviews, By contract access, Dedicated support 【Pros and cons】 Pros: - Speko removes the guesswork of picking a voice model per language by benchmarking 79 models and routing each session to the measured winner, which is valuable for developers who don't have the time or data to run their own evaluations. - The pre-response failover and attempt ID give developers a clear picture of which provider actually served each request, simplifying debugging and reliability monitoring. Cons: - Measured routing is available as public preview access. - Router STT requires an Idempotency-Key on every POST request. - Phone calls add a carrier charge plus 20%, with a minimum of $0.01/min inbound and $0.015/min outbound. 【FAQ】 Q: How does Speko's pricing work? A: Speko offers three plans: Router at +5% on the provider's rate, Speko infra at $0.09 per minute, and Enterprise with custom pricing locked for the term. Q: Who is Speko for? A: Speko is for developers building voice agents and teams whose users speak languages other than English. Q: Does Speko support phone calls? A: Yes, Speko can place phone calls on Speko numbers. Carrier charges are passed through with a 20% markup, at least $0.01/min inbound and $0.015/min outbound.
Router · +5% on the provider's rate (Speech-to-text, LLM, TTS at the provider's published rate, Measured selection and pre-response failover, Public preview access) Speko infra · $0.09/minute (Speech-to-text, LLM, TTS included, Measured selection and pre-response failover, Public preview access) Enterprise · Custom locked for the term (Speech-to-text, LLM, TTS included, Measured selection and pre-response failover plus deployment reviews, By contract access)
WPS AI is an AI assistant built into WPS Office that helps users write documents, generate presentation outlines and…
Baidu Comate is an AI coding assistant built by Baidu on its ERNIE model. It goes beyond simple autocomplete: it…
UniConverter is an AI media conversion and enhancement tool for video, audio, and images. It combines everyday…
6 similar tools to consider instead of Speko