Say it out loud with my Mac's voice model
Say words out loud in a natural voice, the person's own cloned voice or a neutral one, made by a speech model on the person's own Mac, and keep the sound in their files so any of their devices plays it
Runs on the person's Mac (computer/speak.mjs with the speech models of computer/speak.json, such as Qwen3-TTS for a cloned voice and Kokoro for a neutral one, on the Mac's own chip through mlx-audio; nothing is sent to a paid service). Says text in the voice asked for, else the voice of the place it is said in, else the person's chosen voice (rows of their own Voices list and list of places Trillion speaks, read through tool:voice-for), else the shared voices in order, passing over any that cannot be made with why. The sound is kept under /Voice/Said at its natural pace and answered as { file, name, seconds, voice, engine, produced_by, speed }, speed being the place's speed for whoever plays it. tool:speak places this on whichever of the person's devices can run it.
What it asks for
- text (Needed, words): The words to say
- voice (Optional, words): A voice id of the person's (a row of their Voices list); the place's voice, else their chosen voice, when left out
- place (Optional, words): Where this is said: a row id of the person's list of places Trillion speaks, whose voice and speed it uses
Use it from your AI
Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.
Say to your AI: “Use Trillion to say words out loud in a natural voice, the person's own cloned voice or a neutral one, made by a speech model on the person's own Mac, and keep the sound in their files so any of their devices plays it”
What your AI calls: run_tool {"tool_id":"tool:speak-on-this-mac"}