How to Manage Live Captions Language Models in Windows 11 Accessibility Captions Settings

Navigating to Accessibility Captions Settings

Everything else in this tutorial starts from the Live Captions settings page, so it’s worth knowing the route. Check first that your PC runs Windows 11 version 22H2 or later. If the Captions page is missing entirely, run Windows Update before you continue.

1. Press Win+I to open Settings. The Settings⚙️ window opens.

   

2. Select Accessibility from the left-hand sidebar. The accessibility options appear on the right.

   

3. Scroll down the right-hand panel and click Captions under the Hearing section. You should now see the Live Captions configuration page.

That’s it. You are on the page that controls Live Captions, and the later sections all begin here.

If you can’t find Captions, scroll further. The Hearing group sits below the vision options and can be off-screen on a small display.

You can also turn Live Captions on without opening Settings. Press Win+Ctrl+L to switch the feature on or off. Or open Quick Settings (the cluster of network, volume and battery icons at the right end of the taskbar), choose Accessibility, and turn on Live captions.

How to Download and Add New Language Models

Adding a language model lets Live Captions transcribe audio in a language other than your system default. You add these packages from the captions menu. The download needs an internet connection, and on some PCs it needs an administrator account.

If your standard user account blocks system-level additions, Windows asks for administrator credentials or the download fails. Connect to an unmetered Wi-Fi or Ethernet network first, because the files are large.

Windows may also start this process for you. The first time you turn on Live Captions, it prompts you to download the speech model for your display language. You can add more languages later from the gear icon in the captions window, under Captions › Language.

1. Open Settings and go to Accessibility › Captions.

   

2. Turn on the Live Captions toggle at the top of the page if it is off. If Windows asks to download the⬇️ speech model for your display language, allow it. The floating captions bar should appear at the top of your screen.
3. Open the captions settings. On most builds, click the Settings (gear or three-dot) icon in the top right corner of the Live Captions floating window. On some builds the same options are on the main Captions page instead.

   

4. Select Add language from the menu. The wording varies slightly between builds, so look for the language entry if you don’t see this exact label. A list of available languages opens.
5. Browse or search the list, then click the language you want to install.
6. Click Download. The model files start downloading⬇️, and you can watch progress in the captions window or settings page.

You should now have a new language model installed. When the download finishes, Live Captions can transcribe audio in that language.

If a language you expect is missing, the likely causes are hardware that can’t run it locally or a Windows 11 build that doesn’t include it yet. Installing the latest Windows updates is the first thing to try. To undo the change, remove the language as described in the storage section below.

Updating Existing Language Models for Better Accuracy

Updated language models improve accuracy, including punctuation, vocabulary and speech recognition. Windows normally installs these updates automatically through Windows Update. You can also check for them manually, which helps if captions are misreading words or a new language feature hasn’t shown up yet.

1. Open Settings and click Windows Update🔃 in the left-hand menu.
2. Click Check for updates🔃 to scan for available packages. Let any updates that appear finish installing💿.
3. Select Advanced options, then click Optional updates. If you see no entry for language model updates, nothing is waiting for you.
4. Tick the box next to any language model updates and click Download & install. Windows shows the download progress on the same page.

That’s it. Your language models are now up to date. Some updates ask for a restart. If one does, restart when convenient, and Live Captions will use the updated model afterwards.

Managing Storage and Removing Unused Language Models

Removing system-level components may require administrator privileges.

1. Open Settings and go to System › Storage. Use this page to see how much space is in use.

   

2. To manage the speech and language packs themselves, go to Time & language › Language & region. Depending on your Windows 11 version, some options are also under Installed apps or advanced storage options.

   

3. Find the language you want to remove under Time & language › Language & region and open its three-dot menu. Live Captions models are not listed as separate packs, so you remove them by removing the language.
4. Click the three-dot menu next to it and select Remove or Uninstall. The pack should disappear from the list and the space is released.

You should now have the language removed and its disk space back. To bring a language back, repeat the download steps from the earlier section.

Advanced Configuration: Forcing Offline Mode for Language Models

If you don’t want spoken audio or transcribed text sent to the cloud, adjust your speech privacy settings. Live Captions is designed to process audio on the device, using the local models you installed. Turning off online speech recognition makes sure other speech features don’t fall back on cloud services either.

1. Open Settings and go to Privacy & security › Speech.
2. Turn off the Online speech recognition toggle. It should now read Off.
3. Return to Accessibility › Captions and confirm that the language models you installed are present and selected for your captioning sessions.

That’s it. Online speech recognition is off, and Live Captions keeps working from the local models. To undo this, go back to the same page and switch the toggle on again. No restart is needed.

Optimizing NPU Hardware Acceleration for Faster Transcription

On PCs with an NPU, Windows 11 hands the AI work for captions to that dedicated chip. This speeds up transcription and uses less battery than running the same work on the CPU. Windows routes these tasks automatically when it detects compatible hardware, so there is no switch to turn on.

What you can do is keep the system current. Install the latest drivers and firmware through your computer manufacturer’s support utility, and keep Windows Update up to date. Outdated NPU drivers are the most common reason a capable PC runs captions slowly or can’t offer a language that its hardware should support.