| # Upstream Source | |
| This repository is a Voice Scribe **mirror** of an upstream model. The model | |
| weights and code are unchanged from the upstream at the recorded revision. | |
| | Field | Value | | |
| | --- | --- | | |
| | Upstream repo | `ggerganov/whisper.cpp` | | |
| | Upstream revision (sha) | `5359861c739e955e79d9a303bcbc70fb988958b1` | | |
| | Mirror created | `2026-05-07 18:21:41 UTC` | | |
| | Mirror slug | `whisper-metal` | | |
| | Description | Whisper large-v3-turbo - whisper.cpp GGML q5_0 payload for Metal runtime. | | |
| ## Why mirror? | |
| The shipping Voice Scribe installer pins every Apple model to the `VoiceScribe/*` | |
| namespace for a single source of truth, integrity check, and future CDN | |
| migration. Upstream repos retain their original license (see | |
| `LICENSE*` / `README*` files preserved unchanged below). | |
| ## Maintenance | |
| When upstream publishes a new revision we want to adopt, run the matching | |
| Voice Scribe mirror script with `--only whisper-metal` from the repo root. The script | |
| creates a new commit on this mirror that replaces the snapshot and updates | |
| this `UPSTREAM_SOURCE.md`. | |