Re: State of speech-to-text in Debian?

Hi,

在 2026/3/25 8:56, Otto Kekäläinen 写道:
> Hi,
> 
> I noticed Fedora 42 had this release goal:
> https://fedoraproject.org/wiki/Changes/ibus-speech-to-text
> 
> What is the current state of speech-to-text input method / apps in
> Debian? Anything you recommend?
> 
> I tried to find information on the topic, but I didn't find anything,
> which could mean there is no speech-to-text support in Debian
> currently. The closest thing I found was an RFP: vosk-api (#1051812)
> from 2023 but no activity recently.
> 
> In the mobile Linux world there is also
> https://keyboard.futo.org/voice-input-models which is based on
> Whisper, and maybe if those models were in Debian something in Gnome
> could re-use them.
> 
> I might have energy to package something, but first I need to get a
> better understanding of the landscape in general and what is the state
> in Debian currently around this topic.

I am not aware of any that has been packaged. One major issue is that many
contemporary speech-to-text solutions require neural models to work
properly, which may have dfsg issues when packaged in Debian.

A more possible example that may integrate with existing input methods could
be https://github.com/xifan2333/fcitx5-vinput .

Thanks,
Boyuan