Improve Multilingual Dictation UX with Profiles and Quick Profile Switching

[okanakyol]

As a multilingual speaker, I run into some issues with automatic language detection. When Handy/models try to detect the input language automatically, they sometimes choose the wrong language and produce unrelated text. However, when I explicitly set the input language, transcription is much more stable.

I also have a few different daily workflows:

• Speak in my native language and transcribe in that same language
• Speak English and transcribe in English
• Speak in my native language and translate the output to English
• Use auto-detect when I do not care about forcing a specific language

Right now, these workflows are controlled by a single language setting plus “Translate to English”, which makes switching between them a bit cumbersome.

Add Dictation Profiles for common transcription scenarios, and allow switching between them quickly.

Example profiles (in my use cases):

• Auto Dictation
• Turkish Dictation
• English Dictation
• Turkish → English

Each profile would define:

• Input language
• Whether translation to English is enabled
• Potentially, post-processing behavior in the future
  ... those can continue, but in my use-cases those are enough.

I quickly vibe-coded this locally, and it works well for my use cases.

I added a profile section to the General tab (see video):

Screen.Recording.2026-06-21.at.17.56.07.mov

I also added a keyboard shortcut, Option + Shift + Space, to quickly cycle through profiles (see video):

My.Movie.mp4

Curious what others think about this approach.

[alber70g]

I have the same question, only I would like to use two different models and then I wouldn't want to switch with a profile but want to activate with a shortcut. So the one shortcut translates not and the other translates to English.

[VitoMakarevich]

Hi, everyone, I know that new features are not accepted, so I would open a PR.
Perhaps if my fork are being used by more people/I receive reactions here - I can raise PR and point these comments as evidence of demand.
In a nutshell - I made existing post-processing tab - a list of profiles - so what was by default single post processing - became just default tab.
So you can fully customize post processing as before and assign a separate keybinding into each one. e.g. I use local/gpt-6-luna/gpt-6-sol on f5/f6/f7.
The build is compatible with current main branch - your existing settings just migrate into Default profile.
You can read more in my fork/use it. Attaching a demo as well