Toggle punctuation and capitalization independent of STT model
Description:
VoiceInk currently always applies punctuation (e.g., trailing periods) and capitalization (e.g., initial uppercase letters) in transcriptions. I did not find a setting where I can switch it off, I'm sorry if I have overlooked it.
I am using VoiceInk with DeepGram's Nova (the only ~free model currently available to on my device, since local models are not supported on intel macs).
This behavior is problematic when I only want to inject a short phrase, half a sentence or a single word.
In such cases, the automatic capitalization and punctuation feel more like bugs than features.
Feature Requests:
-
Automatic behavior: Ideally, VoiceInk could detect whether the transcription is a full sentence or a fragment, and apply or strip punctuation and capitalization accordingly. This might be part of post-processing, of course speed should not be strongly compromised.
-
Manual toggle: A simpler, immediate solution would be an option to toggle automatic punctuation and capitalization on/off in the UI or settings. (Apologies if this already exists—I couldn't find it.)
Use Case Example:
- I dictate the phrase:
"with automobile"Current output:With automobile.Desired output:with automobile
Why this matters:
VoiceInk is great for quick text injections in middle-of-code or markdown workflows. In those cases, automatic punctuation/capitalization is disruptive and requires manual clean-up.
Environment:
- VoiceInk version: 1.6
- Speech model: DeepGram Nova (cloud)
- OS: macOS Seqoia, Intel mac
Thanks for considering!
Oh yes, please. Something like this would make voiceink so much more usable for editing text.
This is similar to something I proposed and have written code for and I'm willing integrate into with approval, please look athttps://github.com/Beingpax/VoiceInk/issues/380#top
This would be really useful. I just started using the app and I would like to have an option to dictate a short phase and stop. Then paste something from clipboard and then start dictating again. Or type out text partially and then finish everything with the dictation