# Module 3 media provenance and privacy

All 40 equipment/flow/chart illustrations are generated from original Python drawing code in academy/module-3/render.py. No third-party photographs, logos, music, patient/client records or personal voice cloning are used. Video is H.264/AAC; source graphics are PNG. Stock Kokoro af_heart voice is synthesized with the publicly released ONNX v1.0 model/voice bundle. The model publisher declares Apache-2.0; the wrapper is MIT. Notices are retained. Observed model/voice hashes and installed runtime version are recorded by production in voice-provenance.json. No claim is made to independently verify upstream training-data rights or exclusive generated-output rights.

Publisher: https://huggingface.co/hexgrad/Kokoro-82M
Runtime: https://github.com/thewh1teagle/kokoro-onnx
Bundle: https://github.com/thewh1teagle/kokoro-onnx/releases/tag/model-files-v1.1
Font: DejaVu Sans/Bold; retained notice assets/academy/dejavu-font-notice.txt. Web pages use system fonts.

No paid voice API or external media streaming host is required. Generation runs in GitHub Actions; final assets are committed and served from the website. Model weights and voice vectors are not redistributed in the repository. The final repository assets are portable, not dependent on expiring media URLs.

Video requests follow Play; page/image/caption requests may occur earlier. No module-added analytics or third-party embeds. Optional localStorage valistera_module3_v1 contains consent preference, studied/passed lesson IDs and latest score. Workbook text and individual answers are not stored/transmitted. Opt-out/Clear remove only this module key. This is not an audited LMS record. Hosting-level logs, retention, deletion and account settings remain operator responsibilities, not independently verified by this package.