Getting Started ​

Open source speech-to-text for the terminal: Linux-first, provider-agnostic, and scriptable.

OSTT records from your microphone, transcribes with the provider and model you choose, keeps local history, and works from both the shell and a global hotkey popup. It does not assume one vendor, one subscription, or one app-specific workflow: bring your own API key and choose from OpenAI, Deepgram, Groq, DeepInfra, AssemblyAI, Berget, ElevenLabs, or local Whisper-compatible models.

OSTT is built for people who treat the terminal as a normal place for voice input to land. You can print to stdout, copy to the clipboard, write to files, retry the same recording with another model, transcribe existing audio, and post-process text with AI prompts or shell commands. Voice becomes text that can move through the same tools as everything else.

Core Features ​

• Linux-first voice input - Global hotkey setup for Omarchy/Hyprland, GNOME, KDE, and other Linux desktops, with macOS support too.
• Provider choice - Bring your own API key and switch between OpenAI, Deepgram, Groq, DeepInfra, AssemblyAI, Berget, ElevenLabs, and local models.
• Local transcription - Download curated local models or add custom Hugging Face/direct model files for offline transcription.
• Terminal-native workflow - Use stdout, clipboard, paste output, files, aliases, shell completions, logs, and pipes.
• Scriptable post-processing - Transform transcripts with AI prompts or bash commands using ostt -p and ostt process.
• Retry without re-recording - Save recordings locally, then re-transcribe them with a different provider or model.
• File transcription and replay - Transcribe existing audio files and replay saved recordings from history.
• Keywords and text replace - Improve recognition with keywords, then fix final casing, acronyms, and product names with deterministic replace rules.
• Open source, no subscription - Public code, local configuration, and no vendor lock-in beyond the providers you choose.

Install ​

curl -fsSL https://ostt.ai/install | bash

The installer detects your platform, installs supported runtime dependencies, downloads the latest release, verifies its checksum, and installs the ostt CLI.

See the Installation page for alternative install methods and per-platform dependency setup.

Choose A Model ​

Run the model picker to choose the transcription model OSTT should use:

ostt model

Choose Cloud provider for hosted models or Local provider for offline models. Local models can be downloaded and activated from the same screen. If you are unsure which path fits your use case, see Choosing a Transcription Model or Provider.

Cloud providers require credentials first:

ostt auth login

API keys are stored separately from the main config file at ~/.local/share/ostt/credentials with restricted permissions (0600). See Local Models for offline setup.

Record ​

ostt

During recording, the terminal shows:

• Visualization: Real-time audio display (spectrum or waveform, configurable)
  • Spectrum mode (default): Frequency distribution across the voice range (100--1500 Hz focus), with noise gating that suppresses background noise automatically
  • Waveform mode: Amplitude envelope over time, classic oscilloscope-style display
• Vol %: Current volume level
• Peak %: Maximum volume in the last 3 seconds
• Red indicator: Clipping warning when volume exceeds the configured threshold

Controls ​

By default, the transcription is printed to stdout.

ostt -c              # Copy to clipboard
ostt --paste         # Paste into the focused app
ostt -o notes.txt    # Write to file

Use a Global Hotkey ​

Bind this command in your desktop environment or macOS Shortcuts.app:

ostt launch --paste

The first hotkey press opens a popup terminal and starts recording. Pressing the same hotkey again sends a signal (SIGUSR1) to the running OSTT process, which stops recording and transcribes.

Use ostt launch -c if you prefer to copy the result and paste manually. Use ostt launch --paste when you want the transcript inserted into the app that regains focus after the popup closes.

On Linux, paste shortcuts differ between apps. OSTT defaults to shift+insert on Omarchy, ctrl+v on other Linux desktops, and cmd+v on macOS. If your target app needs a different shortcut, set [output.paste].paste_key in ~/.config/ostt/ostt.toml.

See Platform Setup for macOS, Omarchy/Hyprland, GNOME, and KDE setup notes.

Process the Result ​

Processing actions transform transcriptions after recording. For example, this records, transcribes, runs the clean action, and copies the processed output:

ostt -p clean -c

See Processing Actions for action configuration and examples.

Fix Casing and Product Names ​

Use ostt replace when the model hears a term correctly but formats it wrong:

ostt replace

Rules are saved in ~/.config/ostt/ostt.toml under [text.replace]:

[text.replace]
"ostt" = "OSTT"
"api" = "API"
"github" = "GitHub"

Replace runs before processing actions, output, and history saves, so downstream commands receive the fixed text.