[Feature Request] Trigger and send transcript to Shortcuts

[DeviousDroid]

Apologies for posting this as a bug report, but attempting to post this a feature request wasn't working.

Firstly, congratulations on this fork. It's really quite slick.

I was wondering if it was possible to somehow send the transcript to Shortcuts, because doing so would allow one to use Private Cloud Compute to polish the transcript.

I have rigged together a system whereby I process the transcript using the on-device model to add the word TRANSCRIPTION: to the front of the transcript. I use another app that monitors the clipboard for changes and triggers a Shortcut. And in Shortcuts if the clipboard text starts with TRANSCRIPTION:, it uses Private Cloud Compute and my prompt to polish the text. It's over the top, but it does actually work.

If instead Handless could trigger a Shortcut, and pass the text along, it would simplify the process considerably.

In my testing Private Cloud Compute is pretty good for this.

Thank you.

[ElwinLiu]

@DeviousDroid thanks for bringing that up, however, I'm not quite sure why we need Apple Shortcut here, if I'm understanding correctly. Why not set up llm provider inside handless?

[DeviousDroid]

Hi, for a few reasons.

1. Sending the transcription to Shortcuts would allow one to use Private Cloud Compute, which is actually OK for small clean-up tasks and post-processing.
2. Private Cloud Compute is private and pretty fast.
3. It would mean not having to run a model locally – I believe Private Cloud Compute is larger than most users can install locally.
4. It could display a Shortcuts prompt to ask the user to pick from a pre-made set of tasks to tell Private Cloud Compute how to process the transcription, instead of having to select one beforehand.

Those are my thoughts as to why it would be useful. No one is using Private Cloud Compute by sending tasks via Shortcuts, especially as the local model is very limited.

[ElwinLiu]

Hey @DeviousDroid, thanks for the detailed explanation!

Just wanted to let you know — Handless actually already has Apple Intelligence built in as a native post-processing (polish) provider. On macOS Apple Silicon devices, you can select it directly in Settings → Polish → API Configuration — no API key needed, no cost, and it stays within Apple's privacy framework (Private Cloud Compute). So points 1–3 from your reply should already be covered today.

That said, I think the real gap might be point 4 — the ability to dynamically pick a prompt after recording, rather than having to bind a specific prompt to a shortcut beforehand. Is that the main missing piece for your workflow?

If so, we could look into supporting that natively within Handless (e.g. a prompt picker that pops up after transcription) — which would be simpler than routing through Shortcuts. Would that work for you?

[DeviousDroid]

Hi @ElwinLiu are you sure about the Apple Intelligence setting? I'm familiar with the setting, and as far as I'm aware, the only Apple Intelligence available to developers is the smaller on-device model, with Private Cloud Compute only available to Apple and to users via Shortcuts. My understanding is that developers cannot access Private Cloud Compute.