Kekoso

Speech to text on a Mac usually means paying three or four services to cover it: one for dictation, one for meetings, one for converting files, an API key for everything else. Kekoso does all four, and the recognition runs on hardware you already own. Dictation. Press a hotkey, speak, and the text appears wherever the cursor already was, be it Mail, Slack, a code editor or a browser form. No per-app integration, no window to switch to, no waiting on a server. Files and links. Drop in MP3, M4A, WAV or MP4, a voice memo off your phone, or paste a link to a video you have the rights to. You get a transcript with timestamps that exports to TXT, SRT or VTT. Calls. Record from your own side and get the transcript split in two, your microphone on one track and everyone else on the other, so nobody has to admit a notetaker bot to the meeting. Agents. A local MCP server with eight tools lets Claude Code, Claude Desktop and Codex transcribe through Kekoso and manage vocabulary and models, without an API key or per-minute billing. Five speech models are downloaded on demand and run locally: two Parakeet models for speed, Whisper small and large-v3-turbo for reach across up to 99 languages, and SenseVoice for Chinese, Cantonese, Japanese and Korean. A custom vocabulary handles names and jargon that general models get wrong. There is no account, no sign-in and no telemetry anywhere in the recognition path, so the app works with no network at all and keeps working if the vendor disappears. macOS 14.4 or later, Apple silicon. 29 dollars once for three devices, seven-day trial, no subscription.