Feature Request: Capitalization Options

[Hate9]

I love the voice input, but I basically can't use it because it doesn't output text that looks like I wrote it! This is because, in almost all contexts (with the notable and obvious exception of Github issues), I don't capitalize (most of) my words. Currently, FUTO Voice Input doesn't have any options at all to disable auto-capitalization, and this is very annoying.

Many people prefer to not capitalize certain parts of speech which are capitalized in "correct" (standard classroom) English, and FUTO Voice Input should be able to reflect this.

A basic option would be great, but a slightly more in-depth set of advanced capitalization options would be excellent. Here's what I have it mind:

• Automatically capitalize words (y/n) (in the normal input settings)
• Expanded options in the advanced input settings:
  • Capitalize first word of sentence (y/n)
  • Capitalize proper nouns (y/n)
  • Capitalize "I" and contractions of it ("I'm", "I've", etc.) (y/n)
  • Capitalize acronyms (y/n)
  • Capitalize first word of input (y/n) (some people seem to care about this in addition to whether or not the first word of each sentence is capitalized, so we may as well have the option)
  • Capitalize Every Word (y/n) (look, it looks weird to me too, but some people do it, and we may as well include the option. good software doesn't unnecessarily restrict how people interact with it)

This should be relatively easy to implement with some simple post-processing, and should not require tinkering with the transcription model itself.

This ticket relates to #35 and #5.

[jonahbrawley]

I also would love these suggestions to be implemented, I constantly am removing the trailing period from short sentences (such as "I'm on my way.") since it seems unnatural.

There have been other issues suggesting this in the past, and I fail to see why it has been so debated in those threads. The OpenAI Whisper model is really great for voice recognition in this use case but types way too formally for me. If you want every one of your texts to have precise punctuation, nothing wrong with that and more power to you. I personally don't see any harm with this as an optional feature for those who do not.

[laundmo]

Wanted to add that these options to transform the output of the model would be great. Especially the periods at the end of a input are often wrong, as in realtime text chats, linguistically the fact the message is sent counts as the end of sentence marker, and adding a period is doubling up on this, often communicating annoyance or anger

Ideally, i'd love to be able to write my own fix-up rules using something like regex substitutions with maybe a extended syntax like sublime text uses where you can prefix the substitution with \L to lowercase it. But obviously this is quite a expert feature and just having a bunch of predefined toggles would already help a lot.