ChordVox

ChordVox solves a narrower problem than most dictation tools, and solves it more completely. The entire product is built around one loop: hold a hotkey, speak, and the finished text appears wherever your cursor already is. That sounds trivial until you've used a dictation app that transcribes into its own window and leaves you copy-pasting into the real target. ChordVox's auto-paste with native Globe/Fn key support on macOS and true push-to-talk removes that step, and the multi-monitor support means it doesn't break when you work across displays. The privacy story is the real differentiator, and it's structural rather than a marketing claim. Transcription runs on-device through bundled Paraformer, Parakeet, and Whisper/GGML engines, with BELLE Whisper as a Chinese-optimized option — three local engines, 100% local speech-to-text in local mode, zero audio uploads unless you invoke a cloud feature. The community edition is AGPL-3.0 licensed, so the claims are auditable rather than trusted. Cloud transcription and AI polishing are opt-in and explicit, and the vendor states it doesn't collect voice input on its own servers — the only online check is license activation. ChordVox does not require Python; native binaries work out of the box, which matters because local Whisper stacks are usually an afternoon of dependency wrangling. The Pro tier is where the writing actually gets good. The AI polish pipeline takes your raw speech and fixes grammar, punctuation, and paragraphing — the demo on the site turns "Uhh let's do a meeting at four this afternoon, bring Mike, align scope and timing" into "Project sync today at 4:00 PM. Please invite Mike. Agenda: align scope, timeline, and next owners." You connect GPT, Claude, Gemini, Groq through your own API key, or point it at a local GGUF model via llama.cpp if you want to keep the polish step offline too. The multiple shortcut workflows let you assign one key to dictation, another to translation, another to polishing, and another to AI Q&A, each with its own model — that's genuinely more flexible than the single-shortcut model most competitors ship. The custom dictionary is the sleeper feature for anyone with a specialized vocabulary; adding "Parformer, ChordVox, GGUF" to the dictionary means the model prioritizes those terms on future transcription. Where it doesn't fit: there is no iOS or Android app, no team collaboration or shared workspaces, and no cloud sync, so your dictionary and history live on one machine. Handheld dictation is simply out of scope. The AI polish tier requires you to set up and pay for your own model provider — if you want polish that works the moment you install, a managed competitor is a shorter path. And the interface copy is bilingual Chinese/English, which tells you the primary market is Chinese-speaking users; English-language documentation exists but is thinner. Against the field: Wispr Flow and Typeless are cloud-first and offer mobile coverage ChordVox doesn't attempt.