Offline Speech-to-Text on Intel Macs (Metal GPU)
This repository provides a verified, real-world build guide for whisper.cpp running on Intel-based macOS systems with Metal-enabled AMD GPUs (tested with RX 580 8 GB on macOS Sequoia 15.7.2 via OCLP).
The goal: keep legacy Intel Macs fully usable for offline AI workloads.
- ✅ Fully offline – no OpenAI API required
- ⚡ Verified Metal 2 acceleration on AMD RX 580
- 🧩 macOS Sequoia 15.7.2 (OCLP patched) tested
- 🧠 Step-by-step Homebrew + CMake build
- 💾 Model download, quantization (q5_0) and transcription examples
- 🪶 Optional PDF export via Pandoc + Tectonic
Full instructions:
👉 Whisper_on_Intel_Macs_with_Metal_Support_v1.0.md
v1.0 – November 2025
Compiled by Retroniks
MIT License – see LICENSE