Clarification on System Requirements

[nickolasjadams]

FAQ says

Which Macs are supported?
VocaMac works on any Mac running macOS 13 (Ventura) or later. It runs on both Apple Silicon (M1/M2/M3/M4) and Intel Macs, though Apple Silicon provides significantly better performance thanks to CoreML and Neural Engine acceleration.

Up higher in the requirements, it says

Requirements
Apple Silicon (M1/M2/M3/M4)

Tried downloading the dmg on the homepage, but it's telling me no intel.

Just want to clarify if there's intel support or not. If I build from source, will it work?

[jatinkrmalik]

TL;DR: VocaMac is effectively Apple Silicon only right now. Sorry you hit the wall, I'll tighten up the docs so nobody else gets misled. 🙏

@nickolasjadams Hey, sorry for the confusion, that FAQ line is wrong and I'll fix it.

The shipped DMG is Apple Silicon (arm64) only - that's why the installer is blocking you on Intel. The README "Requirements" section is the accurate one, the FAQ on the website is stale and I'll get it updated.

Building from source on Intel might compile (swift build -c release --arch x86_64), but it's not tested or supported, and WhisperKit leans heavily on the Neural Engine — so even if it runs, transcription on Intel CPU will be painfully slow. Honestly, I wouldn't recommend it.

[nickolasjadams]

Thanks for the answer!

[jatinkrmalik]

Quick follow-up — I went ahead and put together an experimental Intel (x86_64) build so you can try it out:

👉 https://github.com/jatinkrmalik/vocamac/releases/tag/experimental-intel-x86_64

A few things to flag honestly upfront:

• It is unsigned and un-notarized, so Gatekeeper will block it on first launch. You'll need to strip the quarantine attribute — full instructions are in the release notes.
• VocaMac uses WhisperKit, which is built around the Apple Neural Engine. On your Intel Mac it'll fall back to CPU-only inference, which is dramatically slower (think 5–10× slower for the same model). Stick with the Tiny model for the first test.
• This is a one-off build from a throwaway experimental/intel-build branch — not a supported configuration. Whether we keep producing Intel builds depends entirely on whether real-world testing shows it's even usable.

If you do try it, I'd love to hear back here on this thread:

1. Does it launch at all?
2. Does the Tiny model download finish?
3. Does push-to-talk actually produce text?
4. Roughly how slow is transcription on your Mac (chip + RAM would help)?

Separately, the doc fixes for the "Intel is supported" misinformation are in #138 — that should land shortly. Thanks again for the nudge — appreciate the patience! 🙏