A new startup called Speko launched today on Hacker News with a clear mission: become the OpenRouter for voice AI. The Y Combinator S26 company aims to provide a unified API gateway for developers to access multiple voice AI models without managing separate integrations.

What You Need to Know

Speko is a new API gateway that aggregates voice AI models from different providers, similar to how OpenRouter aggregates large language models. The startup is backed by Y Combinator’s S26 batch and launched via a Show HN post. Developers using voice assistants, transcription services or speech synthesis can now use a single API to switch between models.

The Voice AI Fragmentation Problem

Voice AI tools are multiplying rapidly. Companies offer speech-to-text, text-to-speech and voice cloning through separate APIs, each with its own pricing and latency characteristics. Developers often lock into one provider and struggle to compare alternatives or migrate when costs rise.

Speko aims to solve this by acting as a middleware layer. The startup accepts requests through a single endpoint, routes them to the user’s chosen model and returns the result. This lets developers test multiple providers side by side without rewriting code.

How Speko Works

The service targets teams building voice-enabled products. Speko handles authentication, rate limiting and billing across providers. Early use cases include real-time transcription for meetings, voice assistants in customer support and content generation with synthetic voices.

  • Unified API: One integration gives access to models from companies like ElevenLabs, Deepgram and OpenAI.
  • Cost optimization: Developers can route low-priority requests to cheaper models and high-quality requests to premium ones.
  • Model switching: Swap providers mid-project without changing application code.

The approach mirrors the success of OpenRouter in text-based AI. OpenRouter became popular by abstracting away the complexity of comparing hundreds of LLMs. Speko applies the same idea to the voice domain, which has historically lagged in standardization.

Why This Matters

Voice AI is entering a phase of rapid commoditization. Model quality is improving faster than pricing stabilizes. Startups that choose the wrong provider early risk expensive migrations or suboptimal performance. Speko lowers that risk by letting developers treat voice models as interchangeable resources.

For larger enterprises, the gateway model could become a procurement tool. Companies can enforce budgets by capping spending per provider or redirecting traffic during outages. Speko also collects usage data across providers, giving customers leverage in price negotiations.

The Y Combinator backing signals investor confidence that the voice API market will grow large enough to sustain a middleman. If Speko gains traction, it could force voice AI providers to compete more aggressively on price and reliability, benefiting the entire ecosystem.