Kai Audio Standard v1.0-beta

Status: transparent synthetic-reference policy

The published beta audio is synthetic reference material. It is useful for practice, coverage checks, and discovering where a spelling-to-sound rule is unclear. It is not presented as a native, human, or final performance of Kai.

Current source of truth

kai_v1_0_beta_pronunciation.md and remain subject to human phonetic review.

text, and human-review state for every shipped beta track.

audio/source_text/.

• IPA and Roman spelling rules are published in
• data/kai_audio_manifest_v1_0_beta.json records the engine, voice, source
• Exact spoken source text stays beside every audio file under

Synthetic audio disclosure

Current tracks use Microsoft Edge TTS with en-US-AriaNeural. The choice is a reproducible draft voice, not evidence that English phonology should determine Kai. A learner should treat IPA, spelling rules, minimal pairs, and later human phonetic review as more authoritative than a model’s accidental accent.

Replacement rule

When a human recording replaces a synthetic file, add the performer consent, recording date, microphone/edit notes, source-text revision, pronunciation QA, and a manifest update in the same change. Do not overwrite a file silently and leave an old transcript pretending it is the same performance.