Replies: 2 comments 1 reply
-
|
I wanted this too, inspired by this article. Here's my draft of one take on this. It seems to work reasonably well on my end.
Working implementation here: main...akurilin:Handy:feat/apple-speech What it does: Adds an "Apple Speech" model backed by the SpeechAnalyzer API on macOS 26+ (Apple Silicon). It shows up in a separate "System Models" section since the OS owns the model assets. There's nothing for Handy to download, store, or delete, it's all managed by the OS. Which deviates from the existing pathway for how models are downloaded and deleted on user's command. Right now it's in a separate section. The Swift bridge follows the same pattern as the existing Apple Intelligence bridge: compiled by build.rs, with a stub fallback when the SDK is older than macOS 26 (you need the Xcode 26 SDK to build the real thing). One quirk worth knowing: SpeechTranscriber punctuates well but starts every sentence lowercase and has no API option to change that, so there's a small post-processing pass that capitalizes sentence starts (unit-tested). I've been using it for real dictation: accuracy is good, punctuation lands where you'd expect, and it runs at roughly 25–40x real-time on my M-series machine. Matches what @apepper saw. Quick 5x run against a 34s recording of my voice:
Caveats: Apple Silicon + macOS 26 only, and building it requires the Xcode 26 SDK (older toolchains compile a stub and the model simply doesn't appear). I know Handy is under a feature freeze, so I'm deliberately not opening a PR yet. If there's interest here and the maintainer is open to it, I'm happy to. Otherwise the branch is there for anyone who wants to build or fork from it. Disclosure: built with heavy Codex Sol and Fable 5 assistance; I reviewed, tested, and have been running it for my needs so far. Some of the changes necessary go beyond my expertise, so any feedback would be welcome here. |
Beta Was this translation helpful? Give feedback.
-
|
You can open a PR for this I'll pull it in @akurilin Just double check if you could if there's an existing one or not. Still feel free to submit just link it as well |
Beta Was this translation helpful? Give feedback.

Uh oh!
There was an error while loading. Please reload this page.
-
With macos 26 and iOS 18.1 Apple introduced a native fully local speech detection (see https://www.macrumors.com/2025/06/18/apple-transcription-api-faster-than-whisper/, https://developer.apple.com/videos/play/wwdc2025/277/ or https://developer.apple.com/documentation/speech/speechanalyzer).
I was able to try them in a different app and apple speech runs quite fast.
It would be awesome if Handy would offer this out-of-the-box for apple builds.
Beta Was this translation helpful? Give feedback.
All reactions