# Transcribe what you say with Voxt

[中文](../transcribe.md)

Transcribe types what you say at the cursor of the app you are in. First [download a local speech model](start.md#download-a-speech-model) in Voxt and allow the microphone and Accessibility.

1. Open **Transcribe** in the Voxt sidebar and check that **Speech model** is a downloaded local model. Start with the default shortcut, Right `⌥ Option`.
2. Open an empty text field and click it to place the cursor. Tap Right `⌥ Option`, say “the meeting is tomorrow at 3 pm”, and tap Right `⌥ Option` again. While you speak, the overlay at the bottom of the screen shows the text as it is recognised.
3. The text appears at the cursor. Press `Esc` to cancel at any time.

<div class="vx-anim" data-scene="dictate" data-lang="en"></div>

To remove fillers and tidy up what you say, choose a [language model you set up](byok.md) on the **Transcribe** page. If a result does not appear in the text field, it is on the clipboard: press `⌘V`, then check where the cursor is and Voxt's Accessibility permission. `⌃⌥V` pastes the last result again. If recognition fails, the recording is kept, and **Retry** in **History** recognises it again. The current shortcut is the one on the **Transcribe** page.

## For agents

> Following this page, check Voxt's local speech model and Transcribe shortcut, open an empty text field, and wait for me to say a harmless test sentence. When I finish, check that the result went in at the cursor. If macOS asks for a permission, let me answer it.