Preferences

The Preferences hold the settings that don’t change with each transcription. Open them with the gear at the top right of the window. The changes are saved automatically.

General

• Appearance: System (follows your system), Light or Dark.
• Interface size: the size of the text and the controls, from 80% to 200%. 100% by default. The change is applied right away.
• Interface language: the language of Audiotext, or System language. See the available languages.
• Date format: how the dates of the transcriptions are shown, in the interface language: short (10/4/26), medium (Oct 4, 2026, the default), long (October 4, 2026) or ISO (2026-10-04). The menu shows each one with an example.
• Time format: Automatic (the clock of the interface language), 12-hour (1:30 PM) or 24-hour (13:30).
• Notifications: shows a notification of the system when a transcription is ready (for a folder, when all its files are, and for a watched folder, each time new files are, in a single notification). On by default. On macOS, they come from Script Editor, and on Windows from Windows PowerShell, so they’re allowed or silenced for those apps in the settings of the system. On Linux, they require notify-send (the libnotify-bin or libnotify package).
• Check for updates automatically: checks for a new version when the app opens and, if there’s one, shows a Version … is available button in the top bar that opens its download page. Pre-releases aren’t offered. On by default.

Providers

The providers of the summaries, the questions and the translations:

• Summary → Provider and Model.
• Ask → Provider and Model: the ones that answer the questions about the transcriptions. They’re the ones of the summaries until you change them.
• Translation → Provider and Model. DeepL and Google Translate have no models to choose.
• Ollama → Server URL: the address of Ollama, http://localhost:11434 by default.

Leave the Model empty to use the default one of the provider. The button next to the provider sets its API key.

API keys

The keys of each service, grouped by what they’re for. Click Set API key… to enter one, or Change API key… to replace it (leave it empty to remove it). The keys that are set show a check. They’re kept in the credential store of your system.

Group	Key	Used for
LLM providers	OpenAI API key	The Whisper API, and summarizing, asking and translating with OpenAI
	Anthropic API key	Summarizing, asking and translating with Claude
	DeepSeek API key	Summarizing, asking and translating with DeepSeek
	Gemini API key	Summarizing, asking and translating with Gemini (from Google AI Studio)
	Mistral API key	Summarizing, asking and translating with Mistral
	xAI API key	Summarizing, asking and translating with Grok
Translation	DeepL API key	Translating with DeepL (keys of the free plan work too)
	Google API key	Google Speech-to-Text beyond the free tier, and Google Translate (Cloud Translation API)
Transcription	Hugging Face token	Identifying the speakers with WhisperX

Caution

Each provider charges for the use of its API, for which Audiotext is not responsible. If OpenAI returns the error 429 with a new key, see Troubleshooting.

Transcription

The options of the engines, in one section each.

WhisperX

Compute type, Batch size and Use CPU. See WhisperX advanced options. On a Mac with Apple Silicon, WhisperX transcribes on its GPU unless Use CPU is on. See On a Mac. On the GPU of the Mac, the hints of Compute type and Batch size say they don’t apply, and Use CPU links to the Supported models.

Whisper API

Temperature and Timestamps of the words. See Whisper API options.

Subtitles

How the subtitles are split: the ones shown over the video and in the Subtitles mode, the exported subtitle files and the videos with subtitles. Each option has a value for the landscape videos and another for the portrait ones (taller than they’re wide, e.g. for phones), which are narrower:

Option	Landscape	Portrait
Max. lines per subtitle	2	2
Max. characters per line	42	24
Chinese, Japanese and Korean: the characters per line in these languages, whose characters are about twice as wide as Latin ones	16	10

The lines of a subtitle are evened out, and the languages written without spaces are broken between characters.

Highlight words underlines each word as it’s said in the .srt and .vtt files. Off by default.

About

The version of Audiotext and links to this documentation, the source code on GitHub and the donation page. Check for updates checks for a new version right away: if there’s one, the button becomes Download and opens its page.