Settings Overview

Open Settings with the Settings button at the bottom of the left sidebar. Manage account in the account menu opens the same dialog on its Account tab. The tray menu has no Settings entry.

The dialog lists ten tabs in two groups in its own sidebar, with the app version printed underneath:

Settings — General, Transcription, Post-Processing, System, Permissions, About
Account — Account, Statistics, Plans and Billing, Data and Privacy

Sign out is on the Account tab. This page covers General, Transcription, Post-Processing, System, Permissions and Plans and Billing. Statistics and Data and Privacy have their own pages: Statistics and Data and Privacy.

General Settings

The General tab contains your most frequently used settings:

Keyboard shortcut — Speakly has exactly one shortcut. The General tab shows its keys next to Recording shortcut, and View opens the editor, which also holds How it behaves. The default is the Right ⌘ key on its own on macOS and Right Ctrl + Enter on Windows and Linux.
Microphone — the input device in use; Change lists every available device.
Languages — the two languages: what you speak and what Speakly types; Change opens the dialog for both.

There is no second shortcut for the assistant. Two quick taps on that same shortcut (within 600 ms) open Agent Mode, and only while Enable assistant is on, on the System tab. Only taps count: holding to talk, letting go and holding again is two separate dictations, never a double-tap.

Transcription Settings

Choose how your audio is transcribed:

Local Model (On-Device)

Uses Whisper running directly on your device. Your audio never leaves your computer. The model runs using your CPU (and optionally GPU if available).

Your API Key (BYOK)

Pick Your API Key under Transcription Engine on the Transcription tab and a Provider grid appears with seven tiles:

Groq — Ultra-fast
OpenAI — Best quality
ElevenLabs — Scribe v1
Google — Cloud STT
Deepgram — Nova 2
Mistral — Voxtral
Custom — OpenAI API, for any endpoint that speaks the OpenAI transcription format

The Google tile is Cloud STT, not Gemini

Google here means Google Cloud Speech-to-Text, and the field is labelled "Google Cloud API Key". The app states it plainly: "Google AI Studio keys do NOT work with Speech-to-Text. You need a Google Cloud Console API key." Create the key in the Google Cloud Console, enable the Speech-to-Text API, then paste it in.

Best Performance

For fastest results, use Groq with your own API key. For maximum privacy, use Local Model.

Post-Processing Settings

Post-processing uses AI to clean up and format your transcribed text. Options include:

Ollama (Local)

Run local LLMs like Llama, Mistral, or Gemma on your device. Install Ollama separately and pull your preferred model.

BYOK Providers

Groq — ultra-fast inference
OpenAI — GPT models
Anthropic — Claude models
Google Gemini — fast & cheap
Mistral — European, GDPR
OpenRouter — access to 100+ models
Cerebras — ultra-fast
Ollama (Local) — runs offline on your machine

Each provider's model list is fetched with your key rather than hardcoded, so new models appear without updating Speakly.

System Settings

System settings control how Speakly behaves on your computer:

Enable sound effects — Play sound when starting and finishing voice recording
Playback when recording — Turn other apps' audio down while you dictate; defaults to Much lower, and Speakly restores the volume you had when the dictation ends, even after a crash mid-recording
Auto-paste text — Automatically paste text into the active application; off means text is only copied to the clipboard and you paste manually
Enable assistant — Process highlighted text with voice commands (e.g., "translate", "summarize"); off means voice only transcribes
Launch at login — Automatically start Speakly when you log in to your computer
Show app in dock (macOS only) — Display Speakly icon in the macOS dock. Changes take effect after restarting the app.
Keep audio recordings — off, audio files are deleted immediately after transcription; on, they are kept, which is also what lets you re-transcribe a past dictation from Recent Activity
Audio acceleration — Speed up audio 2x before sending it to an API engine (smaller upload, roughly half the billed audio-minutes, slightly lower accuracy); the local model always hears the original audio
Redo onboarding — the Redo Setup button resets all configurations and runs the setup again

Recording indicator style

Choose how the recording indicator appears on your screen:

Floating Pill — “Bottom center of screen”: 340 x 100 px, 25 px above the bottom of the work area.
Dynamic Island — “Integrated with notch”: 340 x 37 px at the very top of the screen.

This selector is rendered on macOS only. On Windows and Linux the Floating Pill is always used.

Microphone sensitivity

Microphone sensitivity is on the System tab, not on General. The slider runs from 0.5x to 3.0x in steps of 0.5 and defaults to 2.0x, shown as 200%. Test records three seconds and plays it back with the current setting.

Permissions

Speakly requires certain permissions to function properly:

Keyboard: Required for global hotkeys to work from any application
Microphone: Required to capture audio for transcription
Accessibility: Required for keyboard capture, auto-paste, and for reading the visible text of the front window when Agent Mode runs or a rule asks for screen context
Screen Recording: Required for some features (optional)

Privacy Note

Permissions are only used for their stated purpose. We never record or transmit data without your explicit action.

Plans and Billing

Speakly offers two plans:

Free Trial: 7 days of full access, no card required
Pro (LIFETIME badge): $20, one-time payment

100% Local & Private

The panel's own wording: "Your transcriptions, audio and API keys stay on your device. Our server signs you in, verifies your licence, and counts how many words you transcribed with which engine — never what you said." What that counter contains is spelled out in Data and Privacy.