Deprecate natlink?

[kendonB]

Is anyone using natlink any more? Development seems to have stalled, and it supports a very expensive product when very good free alternatives (e.g. kaldi + nemotron) now exist.

[LexiconCode]

Many do yet. I agree there are plenty of good dictation however very few platforms have accessibility integrations. I'm actually in the middle of overhauling natlink from the ground up to support Python 64 bit. That being said I do hope to replace it in the future.

[kendonB]

I wonder if then a way to make dragonfly more maintainable is to make good on #238 and split off content that faces recognition engines into sub/side packages

[kendonB]

overhauling natlink from the ground up to support Python 64 bit.

@LexiconCode Have you posted your branch anywhere? Happy to pitch in on this

[LexiconCode]

overhauling natlink from the ground up to support Python 64 bit.

@LexiconCode Have you posted your branch anywhere? Happy to pitch in on this

Yeah, I will post a branch sometime tomorrow as I'm working on a few bugs as I type.

[drmfinlay]

I am against splitting up this project. It is maintainable as it is, I have simply been unable to work on it lately. I hope to get back into that over the next month or so. It would be nice to have the Kaldi engine optionally produce formatted dictation output similar to that of the Natlink engine. I have a decent dictation-mode grammar for that somewhere, part of which I had intended to include in Dragonfly. I remember it being pretty usable.

…

On Thu, 26 Mar 2026 14:05:39 -0700 Kendon Bell ***@***.***> wrote: kendonB left a comment (dictation-toolbox/dragonfly#405) I wonder if then a way to make dragonfly more maintainable is to make good on #238 and split off content that faces recognition engines into sub/side packages -- Reply to this email directly or view it on GitHub: #405 (comment) You are receiving this because you are subscribed to this thread. Message ID: ***@***.***>

[kendonB]

I think the best modern options for formatted dictation in dragonfly would be a Kaldi + whisper.cpp combination for CPU-only and Kaldi + Voxtral/NemotronASR if the user has a decent GPU. I will try to get some examples up with the three setups

[LexiconCode]

overhauling natlink from the ground up to support Python 64 bit.

@LexiconCode Have you posted your branch anywhere? Happy to pitch in on this

dictation-toolbox/natlink#226

Sorry it took me so long but it's still a work in progress and needs testing