Adding text to speech voices and speech packs in Windows
Windows reads text aloud with its own voices: in Narrator, in Edge’s Read aloud and in any app that uses the system speech. Each language has its own speech pack, so a voice for another language has to be downloaded first, and that download sometimes hangs at “downloading”.
This guide shows where to add voices in Windows 11 and 10, what to check when the speech pack will not download, and what to use when the system voices sound too flat for a video or an audiobook.
Add a voice in Windows 11 and 10
Open Settings, then Time & language, then Speech. In the section for managing voices, choose Add voices, find the language, select it and confirm; Windows downloads the voices in the background. Another route is Time & language, then Language & region: add the language, or open the options of a language you have, and make sure its text to speech or speech feature is installed. Menu names move between Windows versions, so if one path is missing, try the other. Installing languages may require an administrator account.
When the speech pack will not download
Check that the computer is online and that Windows Update itself works, since language features come through it. A connection set as metered can hold optional downloads back, so try another network or turn that setting off for a while. Restart, then add the voice again; if it still hangs, remove the language and add it back. Downloads of a few hundred megabytes can take a while on a slow line.
Choose the default voice
After the download, pick the voice for apps on the same Speech page, under the voice used for text to speech, and set the speed there. Narrator has its own voice choice in its settings, and on recent Windows 11 versions it can also use downloadable natural voices; the guide to Windows Narrator explains them.
Natural voices for videos and audiobooks
System voices are fine for reading the screen but often sound flat in a voice-over. The free SuperClonica engine runs on Windows with neural voices in 31 languages, works offline after the first model download and saves WAV files you can put in any editor. Ten stock voices come with it, and the catalog adds narrators, announcers and characters you can listen to on their pages first.