{"slug": "agent-voice-kit-local-talk-with-your-agents-whisper-supertonic", "title": "Agent Voice Kit: Local talk with your Agents (whisper+supertonic)", "summary": "Francisco Carlos Serra released Agent Voice Kit, an open-source tool that lets users talk to AI agents like Claude Code, pi, Codex, and Gemini CLI locally, with dictation and optional text-to-speech, running entirely on the user's computer. The kit requires Python 3.11+, ffmpeg, and a GPU for optimal performance, and downloads Whisper large-v3-turbo and Supertonic 3 models from Hugging Face (about 2 GB).", "body_md": "**English** · [Español](/franciscocarloserra/agent-voice-kit/blob/master/README.es.md)\n\n```\nyou speak                  the agent answers              you hear it (optional)\npress F1, talk, press F1   the text lands in the window   the reply is read aloud\n```\n\nPress a key, say what you want, press it again. Your words are typed wherever you were writing, in any app. If you want, the agent reads its answer back to you.\n\n## simplescreenrecorder-2026-09-02_12.44.15_audio_v3_1.5x.mp4\n\nIt all runs on your own computer. Nothing is sent anywhere, and there is no waiting for a server: the text is there the moment you stop talking.\n\nPaste this into your agent (Claude Code, pi, Codex, Gemini CLI):\n\n```\nInstall https://github.com/franciscocarloserra/agent-voice-kit/ following its AGENTS.md\n```\n\n**What you will have to do by hand.** The agent cannot click these for you:\n\n- Mac: type your password for\n`brew`\n\n, assign the two hotkeys in System Settings, accept the Microphone and Accessibility dialogs. - Windows: approve the ffmpeg and AutoHotkey installs when winget asks.\n- Linux: type your\n`sudo`\n\npassword for the packages, assign the two hotkeys in your desktop's shortcut settings.\n\n| key | what it does |\n|---|---|\n| F1 | starts recording; pressing F1 again types your words at the cursor, in any app |\n| Meta+F1 | cancels the recording or silences the agent |\n`/tts` |\nturns read-aloud on or off, in any harness. Persists until you toggle it again |\n\n`python3 voice.py serve`\n\nruns the servers in a terminal, `install-service`\n\nruns them at login, `doctor`\n\nreports what is missing, and `tts speed 1.5`\n\nsets the reading speed.\n\n**You want a GPU.** NVIDIA on Linux or Windows, or any Apple Silicon Mac. On CPU only it works but takes a few seconds per sentence.\n\n**Models are downloaded, not bundled.** Setup pulls Whisper large-v3-turbo and Supertonic 3 from Hugging Face, about 2 GB.\n\n**A few things get installed on your system.** Python 3.11+ and ffmpeg everywhere, AutoHotkey on Windows, xdotool on Linux, current NVIDIA drivers if you have that GPU.\n\n**Text to speech works out of the box, and can get faster.** The default engine needs no build. If you want replies to start instantly, ask the agent to build audio.cpp (about 15 minutes, 5 to 8 times faster).\n\n**macOS will ask for permissions.** Microphone and Accessibility, once. Without them dictation records silence or does not type.\n\n**Read-aloud is off until you turn it on.** `/tts`\n\ntoggles it for every harness and it stays that way. Dictation needs nothing: it types into whatever window has focus.\n\n**Everything is customizable.** Models, voice, reading speed, hotkeys, ports. Just ask your agent to change it; it all lives in `config.json`\n\n.", "url": "https://wpnews.pro/news/agent-voice-kit-local-talk-with-your-agents-whisper-supertonic", "canonical_source": "https://github.com/franciscocarloserra/agent-voice-kit", "published_at": "2026-09-02 16:39:08+00:00", "updated_at": "2026-09-02 16:52:32.074853+00:00", "lang": "en", "topics": ["ai-tools", "developer-tools", "natural-language-processing"], "entities": ["Francisco Carlos Serra", "Agent Voice Kit", "Whisper large-v3-turbo", "Supertonic 3", "Hugging Face", "Claude Code", "Codex", "Gemini CLI"], "alternates": {"html": "https://wpnews.pro/news/agent-voice-kit-local-talk-with-your-agents-whisper-supertonic", "markdown": "https://wpnews.pro/news/agent-voice-kit-local-talk-with-your-agents-whisper-supertonic.md", "text": "https://wpnews.pro/news/agent-voice-kit-local-talk-with-your-agents-whisper-supertonic.txt", "jsonld": "https://wpnews.pro/news/agent-voice-kit-local-talk-with-your-agents-whisper-supertonic.jsonld"}}