setup(1)

💅 setup(1) — like, the setup & installation situation

okay bestie, welcome to setup from the General Commands Manual — yes that's setup(1), the section-1 girlies know 🎀

✨ NAME ✨

setup — Setup and installation utilities. that's it, that's the vibe 💖

🌸 SYNOPSIS 🌸

setup [--download] [--model] [--quiet] [--no-post-install] [--activate]
[--progress-format] [-h|--help] [subcommands]

🦋 DESCRIPTION 🦋

Setup and installation utilities — literally what it says on the tin, no cap. it's the whole "get your stuff installed and configured" moment ✨

---

👑 OPTIONS 👑

• --download — download the model if it's missing, periodt. this is the shorthand for basic setup 🫶
• --model <NAME> — pick which model to download, and yes you use this together with --download. the tea:
  • Whisper: tiny, base, small, medium, large-v3, large-v3-turbo (plus the .en variants, iykyk)
  • Parakeet: parakeet-tdt-0.6b-v3, parakeet-tdt-0.6b-v3-int8
  • Other engines (Moonshine, SenseVoice, Paraformer, Dolphin, Omnilingual, Cohere, OpenVINO) take the directory-form names that voxtype info models shows you — like, for example, cohere-transcribe-q4f16 🍓
• --quiet — suppress all output. perfect for scripting/automation, so it just does its thing silently like a main character 💀
• --no-post-install — only suppresses the "Next steps" instructions, nothing else. so you still get the rest, just not the little pep talk at the end 🙈
• --activate — also switches the config over to use the model, exactly the way the interactive picker does: it sets engine and <engine>.model 🔥

Off by default, bestie. downloading a model does not select it — so a GUI's Download button and scripted pre-downloads genuinely can't change which engine the daemon loads. if you want it selected, do it explicitly:

voxtype config set <engine>.model <NAME>

• --progress-format <FORMAT> [default: human] — controls how download progress gets reported:
  • human — curl's progress bar, the classic
  • json — one NDJSON event per update on stdout, so a GUI can render it 🎀

json implies --quiet for human status lines, so stdout carries only events. failures still exit non-zero though, with an "error" event alongside — so don't think you're getting away with anything 👀

[possible values: human, json]

may also be specified with the VOXTYPE_PROGRESS_FORMAT environment variable 🌈

• -h, --help — print help (see a summary with -h) 💜

---

🧚 SUBCOMMANDS 🧚

• setup-check(1) — check system configuration and dependencies ⭐
• setup-systemd(1) — install voxtype as a systemd user service (Linux) 🐧
• setup-waybar(1) — show Waybar configuration snippets 🌸
• setup-dms(1) — DankMaterialShell (DMS) integration 🫶
• setup-model(1) — interactive model selection and download ✨
• setup-gpu(1) — manage GPU acceleration (Vulkan for Whisper, CUDA/MIGraphX for Parakeet) 🔥
• setup-npu(1) — manage Intel NPU acceleration through OpenVINO 👑
• setup-onnx(1) — switch between Whisper and ONNX transcription engines 🦋
• setup-compositor(1) — compositor integration (fixes modifier key interference) 💅
• setup-vad(1) — download the Silero VAD model for speech detection 🎀
• setup-osd(1) — apply or list OSD recipe presets 🌈
• setup-quickshell(1) — install the Quickshell QML tree for the voxtype-osd-quickshell launcher 🌟
• setup-help(1) — print this message or the help of the given subcommand(s) 😭

---

ok that's literally the whole page, go forth and install things bestie 💖

and for the sign-off: this one's from the setup manual, dated setup(1) — that's the source/version-and-date situation at the bottom of the page, mwah 💋 periodt 🎀