Feature request: Translation mode (e.g. French → English) via separate hotkey

[Flolight]

Use case

I often need to speak in French but have the output in English (e.g. when prompting coding agents or writing in English-language channels). Currently Wispr only supports transcription (same language in, same language out).

Proposal

Add a translation mode alongside the existing transcription mode, where speech in any language is translated to English text. The user could toggle between the two behaviors via a separate keyboard shortcut:

• Shortcut A (existing): Transcribe — French speech produces French text
• Shortcut B (new): Translate — French speech produces English text

This way both behaviors are available without changing settings each time.

Thanks for the great tool! 🎙️

[sebsto]

Thank you for the suggestion.
Do you see this as two different shortcuts (which increase the mental workload when using the tool) or a settings to flip in the Settings windows (less flexible to switch between mode)

[Flolight]

Thanks for the quick reply!

My preference would be two shortcuts, since I switch between modes frequently (e.g. transcribing meeting notes in French vs prompting a coding agent in English). The mental load is minimal when the shortcuts are close to each other.

That said, I can see how others might primarily use one mode and only occasionally need the other. Maybe the best approach is to make it configurable: let the user choose between a dedicated second shortcut or a simple toggle in Settings. That way people who mostly stay in one mode can just flip a setting, while frequent switchers get the two-shortcut workflow.

[tobycheese]

This could be extended a bit. Have a separate hotkey that first does STT (via the existing functionality), and then sends the text to a second LLM to further processing. The spoken text would have to have some instructions at the beginning, such as

• "translate the following text to french: Here comes the actual text to be translated"
• "formulate the following points as an email: Then some rough thoughts would come here"

Then the result would be inserted at the cursor. The tool would still stay simple to use and not evolve in a bloated do-it-all-ai-chatwindow-notepad thing. But of course it is a bigger change... have a secondary set of models, a new flow.

[sebsto]

@tobycheese there is already a possibility to send the text to a second local LLM to enhance the TTS result. These two capabilities would be incompatible.