Speaker

Turn text into speech and get the timeline with it. One call returns the audio AND every word's start and end in milliseconds — enough to cut captions, drive a video timeline or sync a mouth, without a second pass over the audio you just made.

• Word timings are DERIVED from the engine's character-level alignment: no speech engine returns word-level timings, so this one groups the characters on whitespace and says so — in the docs, on the homepage, and in a timings_derived_from field on every response
• A pronunciation map respells the names your script keeps getting wrong ({"Ounie": "Oonie"}) before synthesis, and the timings still come back carrying YOUR spelling, so captions are publishable as they arrive
• 21 voices read live from the engine, three models: fastest, warmest, and one that speaks 29 languages. Models whose timestamp support could not be confirmed are absent from the enum rather than present and unreliable
• Priced per 1,000 characters — 10 credits on the fast engines, 20 on the rich one — because speech cost scales linearly with input length and a flat per-call price is wrong at both ends
• A run that returns audio with no usable alignment settles at ZERO and refunds in full; on the keyless rail it is refused before the payment settles
• Audio is private: every download is a fresh short-lived signature, never a public URL. Listing runs, re-reading one and re-downloading its audio are free forever
• Agent-native from day one: spk_live_ REST keys, an MCP server (works with the Ounie AI Team via ?api_key=), keyless x402 pay-per-call in USDC, and the fleet master key

{
  "mcpServers": {
    "ounie": {
      "url": "https://mcp.ounie.com"
    }
  }
}

{
  "mcpServers": {
    "speak": {
      "url": "https://speak.ounie.com/api/mcp"
    }
  }
}