Skip to content

Security: gabrimatic/kokoro-mlx

SECURITY.md

Security Policy

Privacy by Design

Privacy is a core constraint, not a feature toggle.

  • All processing is local. Inference runs entirely on your Mac via MLX.
  • No network calls except to download model weights from HuggingFace Hub on first use.
  • No telemetry, no analytics, no cloud. Zero data leaves your machine after model download.
  • Text stays on device. Input text is processed locally and never transmitted anywhere.
  • Audio stays on device. Generated audio is written locally or played back on device.

Trust Boundaries

Boundary Trust Level Notes
User text input Trusted Processed locally, never transmitted
Generated audio Trusted Written to local filesystem or played on device
Model weights Trusted Downloaded from HuggingFace Hub (Apache 2.0, by hexgrad), cached locally
Voice style vectors Trusted Bundled with model weights, loaded from local cache

No remote trust boundaries during inference. The only network access is the initial huggingface_hub.snapshot_download call to fetch model weights.

Vulnerability Reporting

Report vulnerabilities responsibly:

  1. Do not open a public issue. Vulnerabilities stay private until a fix ships.
  2. Use GitHub's private vulnerability reporting to submit.
  3. Include:
    • Steps to reproduce
    • Demonstrated impact
    • Suggested fix (if any)

Reports without reproduction steps or demonstrated impact are deprioritized.

Expect acknowledgment within 48 hours.

Out of Scope

These are not considered vulnerabilities:

  • Issues requiring physical access to the machine
  • Adversarial text inputs designed to produce specific audio outputs
  • Mispronunciations or inaccurate prosody

Supported Versions

Version Supported
0.1.x Yes

There aren't any published security advisories