How Handy would it be if Handy used the clipboard as context?

[Oblib]

What if Handy could read the clipboard as additional context when post-processing is triggered? This would turn voice into an instruction on existing text, not just dictation.

Proposed flow

1. Option+Shift+Space → Handy clears clipboard, starts recording
2. User copies text with Cmd+C during recording (optional)
3. User dictates an instruction
4. On release: if clipboard is non-empty, it is included in the LLM prompt alongside the transcription

Use cases

• Dev: copy code → dictate "explain this function in plain English"
• Writer: copy article → dictate "summarize in 3 bullets"
• Translator: copy text → dictate "translate to formal English"
• Student: copy notes → dictate "turn into flashcards"
• Medical: copy a list of diagnoses → dictate "format as a numbered list"

The common thread: voice becomes the command, clipboard becomes the data.

Why it fits Handy

• Minimal change — the clipboard plugin is already used for paste
• No regression — if the clipboard is empty, behavior is unchanged
• Generic — useful across many workflows
• Feature parity with SuperWhisper on this specific point

Questions

• Would you find this useful?
• Should it be opt-in (toggle in settings) or always active when post-processing is enabled?
• Should there be a visual indicator when clipboard content is being included?

Happy to work on a PR if there is interest. Thanks!

[pavanm101]

I would like that feature, I was thinking about that too

[boadij]

+1 for this. My main use case is slightly simpler: use the current clipboard contents as optional context for transcription post-processing.

For example, if I've copied code, a GitHub issue, a person's name, or some technical text before dictating, that context could help the post-processing model correctly resolve names and terms that the speech-to-text model misheard.

I wouldn't necessarily want the clipboard treated as text to transform. Ideally it would just be additional context available to the post-processing prompt, preferably behind an opt-in setting.

This could make contextual correction significantly better without changing the normal dictation workflow.