Speech options

Let Voice choose the best fit.

Voice suggests an option based on your computer. Most people can leave it there. Choose another option only if you want faster results, more accuracy or a smaller download.

01 / Pick an option

Choose speed or accuracy.

OptionApprox. downloadWhat to expect
tiny.en75 MBA quick way to check that Voice is working. English only.
base.en142 MBFast everyday English dictation on smaller computers.
small.en466 MBMore accurate English, with a larger download.
medium.en1.5 GBA good balance if your computer can keep up.
distil-large-v31.5 GBFast and accurate on powerful computers. English only.
large-v32.9 GBThe most accurate option here, but it needs the most space and runs more slowly.

Download sizes are estimates. Results also depend on your computer, microphone, language and how you speak.

02 / Downloads

Voice downloads its speech option separately.

Voice downloads the speech file it needs the first time you set it up. Files come from the whisper.cpp model repository on Hugging Face and are checked before installation.

These speech files have their own licences and terms. You can read the details in the third-party notices.

03 / Already set up?

Your existing downloads still work.

If you used OSWispa before, MorpheOS Voice keeps using the speech files you have already downloaded. You will not need to download them again just because the name has changed.

Voice finds the right place for them automatically on each operating system. Technical file locations are listed in the user guide if you ever need them.

Start simple

Most people can keep Voice’s recommendation.