Skip to content

picks

Voice AI infrastructure and speech engines for developers

A curated list of real-time speech APIs, local transcription models, and voice synthesis engines for software engineers building voice applications.

7 tools · curated by Stork

Building reliable voice interfaces requires orchestrating low-latency audio transport, speech recognition, and expressive text-to-speech synthesis. Engineering teams face tradeoffs between managed streaming APIs that minimize round-trip latency and open-weight models that preserve user privacy through local execution. Choosing the wrong abstraction early in development often leads to costly architectural rewrites when latency spikes or connection edge cases emerge in production. This collection highlights low-level voice infrastructure, real-time orchestration frameworks, and specialized speech engines. Evaluate these tools based on your system requirements—whether you need cloud-based streaming pipelines for turn-based conversation, self-hosted models for offline transcription, or expressive speech synthesis for voice agents.
  1. 1

    Vapi

    🤖 Automate

    Voice AI for developers

  2. 2

    Cartesia Sonic

    🤖 Automate

    Sonic is the fastest human-like voice API

  3. 3

    OpenAI Realtime API

    🧩 Deploy

    Bidirectional streaming interface for GPT-4o Realtime.

  4. 4

    Whisper by OpenAI

    🤖 Automate

    A neural net for speech recognition

  5. 5

    Phonon-2

    🤖 AI Tools

    Phonon-2 is an English-focused speech recognition model from Fermion Research, designed for local, on-device transcription.

  6. 6

    KittenTTS 2

    🤖 AI Tools

    KittenTTS 2 is a speech-generation model that can produce new speech sounding like an original speaker from a short audio recording, featuring in-context voice cloning.

  7. 7

    Minicrow AI

    🤖 AI Tools

    MiniCrow offers AI tools for speech-to-text, text-to-speech, real-time voice AI, and video understanding, all on an OpenAI-compatible API designed specifically for Indian languages. The platform includes various models optimized for real-world applications, including LLM, voice agents, and transcription services.