AI BriefingLaunch HN: Speko (YC S26) – OpenRouter for Voice AI
Key point
Speko is a gateway service that optimally routes various voice AI models based on performance and cost.
Details
Speko is a Router service that integrates various voice AI models (STT, TTS, LLM) into a single API, enabling calls to the optimal model.
Instead of relying on English-centric leaderboards provided by vendors, it determines models based on actual language-specific/use-case-specific metrics. Users can compare model accuracy (WER) and cost per minute ($/min) to select the optimal model.
Key Features:
- Unified Gateway: Access models from various providers such as OpenAI, Google, Deepgram, and xAI through a single Base URL and API key.
- Maintain Existing Workflows: Supports OpenAI API specifications, allowing immediate use in existing frameworks like LiveKit, Pipecat, and Claude MCP by simply changing the hostname and model string.
- Optimized Model Selection: Provides performance and cost data for various latest models, including GPT-4o Transcribe, Qwen3-ASR, Chirp 3, Grok STT, and Nova-3.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Open-source 'VoiceStudio' released with support for local voice cloning, dubbing, and transcription
DeepSeek Releases DSec: Elastic Sandbox Infrastructure for Large-Scale Agentic Training