apple-native-transcribe

name

apple-native-transcribe

description

Transcribe local audio files on macOS 26 or newer with the `apple-native-transcribe` CLI. Use when a task needs Apple's on-device SpeechAnalyzer and SpeechTranscriber workflow for `.mp3`, `.m4a`, `.wav`, or similar local audio, and the CLI may need to be verified or installed with `brew install mweinbach/max-skills/apple-native-transcribe`. Do not use this skill on Linux, Windows, or macOS versions earlier than 26.

Apple Native Transcribe

Use this skill only on macOS 26 or newer.

Workflow

1. Confirm the host is macOS and check sw_vers -productVersion.
2. Stop immediately if the host is not macOS 26 or newer.
3. Check whether the CLI exists with command -v apple-native-transcribe.
4. If the command is missing, install the CLI with brew install mweinbach/max-skills/apple-native-transcribe.
5. Run the CLI with an absolute input path, an absolute output directory, and one output format.
6. Read the JSON manifest from stdout first, then inspect the written transcript file if needed.

Do not invent alternate installation paths. The install command for this skill is:

brew install mweinbach/max-skills/apple-native-transcribe

If Homebrew is unavailable or the install fails, stop and report that clearly.

Primary Commands

Use flag arguments:

apple-native-transcribe --input "/abs/path/audio.mp3" --output-dir "/abs/path/out" --format json

Use positional arguments:

apple-native-transcribe "/abs/path/audio.mp3" "/abs/path/out" json

List supported locales:

apple-native-transcribe --list-locales

Supported formats:

• json for debug-oriented structured output that is not meant to be human-readable
• text for the raw plain-text transcript
• srt

Output Contract

The CLI always:

• writes the requested transcript file into the output directory
• writes transcript.debug.json beside it
• prints a JSON manifest to stdout

Format expectations:

• json writes a debug-style structured transcript artifact intended for tools or inspection, not normal human reading
• text writes the raw transcript as plain text
• srt writes subtitle output

The manifest includes:

• transcriptFile
• debugFile
• format
• localeIdentifier
• segmentCount

Prefer json when another tool or model will consume the transcript. Prefer text when a human needs the plain transcript.

Debug Flags

Only use debug flags when the default flow is insufficient:

• --debug-locale <locale-id>
• --debug-install-assets
• --debug-prewarm
• --debug-preset <transcription|alternatives|timeIndexed|progressive|timeIndexedProgressive>
• --debug-context <phrase>

If the CLI reports missing on-device speech assets for the selected locale, retry with --debug-install-assets.