voiceio vs Voxtype
Voxtype and voiceio are both free, MIT-licensed, local dictation tools for Linux with no cloud and no telemetry by default. Voxtype is a Rust binary with nine speech engines, GPU builds for AMD, NVIDIA and Intel, a meeting mode and a macOS beta. voiceio is a Python tool built on faster-whisper that streams a live, corrected-in-place preview through IBus and can fine-tune its model on your own voice. If you want the widest choice of engines and hardware, Voxtype is the stronger pick; if your errors are names and jargon you say every day, voiceio's fine-tuning is the difference.
Which should you use?
Choose Voxtype if…
- You want many engines: Whisper, Parakeet, Moonshine, SenseVoice, Cohere Transcribe and more, including CJK and 1600+ languages through its multilingual engines.
- You have a GPU to use: Vulkan, CUDA 12/13, MIGraphX for Radeon, OpenVINO for Intel NPUs.
- You want a single native binary from AUR, .deb, .rpm or Homebrew, with no Python.
- You need meeting transcription with speaker attribution and SRT/VTT export, or a macOS build (beta).
- You prefer push-to-talk (hold the key, release to type).
Choose voiceio if…
- The same names and terms keep coming out wrong:
voiceio learnfine-tunes the model on your own recordings and only switches when it measurably wins. - You want to see words as you speak, as an underlined preview corrected in place rather than typed and backspaced.
- You want vocabulary hotwords ranked by how often you use them, plus correction rules and a "correct that" voice command.
- You want to dictate from your phone through a server you run, with the same vocabulary and model.
Feature comparison
| voiceio 1.5 | Voxtype | |
|---|---|---|
| Platforms | Linux: Wayland and X11 (GNOME, KDE, Hyprland, sway, i3). Untested best-effort path on Windows/macOS | Linux (Hyprland, Niri, sway, River, GNOME, KDE, X11); macOS beta |
| Price | Free, MIT | Free, MIT |
| Where speech is decoded | On your computer. Optional text-only cloud cleanup, off by default | On your computer by default; optional remote Whisper server or Soniox cloud backend |
| Speech models | faster-whisper small (default), medium, large-v3-turbo, distil-large-v3; experimental Parakeet and whisper.cpp server | Nine engines: Whisper (whisper.cpp), Parakeet, Moonshine, SenseVoice, Paraformer, Dolphin, Omnilingual, Cohere Transcribe, OpenVINO Whisper |
| Open source | Yes (MIT) | Yes (MIT) |
| Live typing into apps | Yes: underlined preview in the focused app through IBus, corrected in place, committed when you stop | Optional streaming mode types at the cursor as you speak (toggle activation; experimental revision mode corrects with backspace). Default: type after release |
| Custom vocabulary | Hotwords ranked by use (about 35 per decode) plus find-and-replace corrections | Whisper initial prompt, replacement table, spoken punctuation, post-processing command |
| Learns from you | Yes: optional LoRA fine-tune on your own recordings, on CPU, kept only if it wins on held-out clips | Not stated |
| Voice commands | Dictation commands only ("new line", "scratch that", "correct that") | Spoken punctuation; no computer control stated |
Voxtype facts are from its official site, docs or repository, read on 2026-10-10 (sources below). Prices are as listed on that date and can change. voiceio facts are from its README. Spotted something out of date? Open an issue.
Questions
Is voiceio a Voxtype alternative?
Yes. Both are local, open-source dictation tools for Linux under the MIT license. voiceio adds a live IBus preview and per-user fine-tuning; Voxtype offers more engines, GPU builds, meeting mode and a macOS beta.
Which is faster?
It depends on engine and hardware. Voxtype's README reports Cohere Transcribe at 9-11x realtime on a Zen 4 CPU and supports GPU backends. voiceio's default whisper-small runs at about 5x realtime on a laptop CPU. Measure on your own machine.
Do both work on Wayland?
Yes. Voxtype types with wtype, dotool or ydotool and uses compositor keybindings. voiceio uses an IBus input method first, with wtype and ydotool as fallbacks.
Does Voxtype learn my voice?
Its README and docs (checked 2026-10-10) do not describe fine-tuning or voice adaptation. It offers an initial prompt and replacement rules for domain terms.
Try voiceio
Free and MIT licensed. Install it with pipx on Linux, or try the dictation in your browser first: the demo on the home page runs a small speech model inside the tab.