[Feature Request]: Voice typing

[thewh1teagle]

Describe the feature

https://github.com/thewh1teagle/vibe/tree/feat/voice-type

New tab with option to select language and microphone and button to start
When active there's global keyboard shortcut (Ctrl + J). while pressing it will record from mic, when stop it will transcribe and type the result.

Components:

• Tauri keyboard shortcut plugin for detect hotkeys
• Enigo for keyboard emulation
• Samplerate for resample to 16000
• Ringbuffer to hold the audio samples while recording
• Ebur128 for normalize audio loudness

What's need to be done:
Load whisper context correctly
Start record
Normalize loudness
Normalize samplerate
Transcribe and type

Questions:
Global indicator of start/stop needs to be somewhere. by hearing or by seeing.

[thewh1teagle]

[Y-PLONI]

Thank you very much! I've been waiting for this!

[ova13lastar]

+1

[Y-PLONI]

@thewh1teagle
Is it ready, or not yet?
[It won't help me yet, because I don't have powerful enough hardware...
But I'd really like to know! I'm really curious!]

[thewh1teagle]

Is it ready, or not yet?
[It won't help me yet, because I don't have powerful enough hardware...
But I'd really like to know! I'm really curious!]

I decided that another tab for app will make it less intuitive and I still want to keep the regular recording behavior that works now so I need to redesign this (both functionality and interaction)

[kyoukisu]

Still holding out for this feature 👉👈

[TomFrankly]

If this will work like the Dictation feature in MacWhisper – inserting the transcribed text directly into the active text field in any app – I'll be eternally grateful. I'm trying to find a Whisper-based solution for this problem for a friend who injured their arm and only works on Windows computers.

[pepijndevos]

Maybe you could add the hotkey to the recording tab and make a toggle for automatically copying the text to the clipboard?

[pkrolkgp]

Is this feature still in progress? I would love to see it in action!

[Angroid]

If this will work like the Dictation feature in MacWhisper – inserting the transcribed text directly into the active text field in any app – I'll be eternally grateful. I'm trying to find a Whisper-based solution for this problem for a friend who injured their arm and only works on Windows computers.

I concur, I've been looking for something like MacWhisper for Windows as I need to work on a Windows computer from time to time and I'm dearly missing it. I found two solutions:

1. Whispertype
   Works as desired and really well, but not possible to configure anything. It also uses online models exclusively but I want offline processing for privacy reasons (I don't believe any LLM provider when they say they don't keep / use my data). It does feel most similar to the MacWhisper dictation feature though. It's free for now but it's only a matter of time until the API costs need to be monetized by the developer.
2. Speechpulse: This app meets all the requirements, but it's 99$.

There a literally dozens of apps out there that are based on Whisper, but most do not provide the dictation / output to any textfield feature of MacWhisper.

I'll end up paying for Speechpulse if I can't find anything better or cheaper. I very much prefer the design of Vibe though and this would make the package perfect. Not to criticize Speechpulse, it's an amazing piece of software with a lot of configuration and complex added functionality. But I only need it for dictation and sometimes transcribing a video which doesn't justify the price for me.

I find it quite surprising that after all those years since the first Whisper model (which was honestly already much better than any other speech recognition) this is still not a core integration in Windows or Apple OS'es. Apple's dictation is not bad (Windows dictation is criminally bad) but it's quantum leaps from what Whisper can provide in accuracy and contextual interpretation.

Edit: I just realized who I replied to. Hey there, keep up the good work, thanks to you my colleagues think I'm a Notion magician :D !

[Angroid]

Maybe you could add the hotkey to the recording tab and make a toggle for automatically copying the text to the clipboard?

If the live typing is technically challenging, automatically copying to clipboard could get us 95% there without much hassle.

[pepijndevos]

I may have to hack this together myself with https://pypi.org/project/whisper-cpp-python/ and https://pypi.org/project/pynput/