audio / production

recording replacement checklist

Use this checklist when replacing synthetic Kai reference audio with human recordings. The current engine and source text are disclosed in the beta audio standard; a track is not ready to replace the public file until the script, recording, edit, and QA gates are complete.

course hub proficiency scale

inventory

tracks to replace

sounds

Core sound inventory

Vowels, consonant rows, syllable practice, and minimal pairs.

track id: sound_inventory
synthetic reference audio: current file
script: 14 transcript lines

word recognition

Core 1,000 word audio

High-frequency Common Kai words with teaching pronunciations.

track id: core_1000
synthetic reference audio: current file
script: 1001 transcript lines

beginner dialogue

Beginner dialogue audio

Short call-and-response exchanges for first-course listening practice.

track id: beginner_dialogues
synthetic reference audio: current file
script: 421 transcript lines

connected speech

Intermediate dialogue audio

Longer Common Kai turns for everyday and story-level listening.

track id: intermediate_dialogues
synthetic reference audio: current file
script: 421 transcript lines

advanced dialogue

Advanced dialogue audio

Dense Common Kai dialogue for modality, argument, and translation practice.

track id: advanced_dialogues
synthetic reference audio: current file
script: 221 transcript lines

sacred register

Sacred Kai sample audio

Short sacred-register samples for careful listening and register comparison.

track id: sacred_samples
synthetic reference audio: current file
script: 4 transcript lines

workflow

recording stages

script lock

Confirm the source text matches the public transcript for every line.
Assign stable track IDs, speaker labels, and take numbers before recording.
Mark pronunciation risks: vowel groups, first-syllable stress, rin-te, an-vai, and sacred-register items.
Freeze the text before recording; any later wording change starts a new take.

speaker prep

Review the pronunciation reference and one model reading before recording.
Keep every Kai vowel clear; do not reduce unstressed vowels to English schwa.
Read dialogue turns with steady pacing and no theatrical overacting.
Record a short calibration sample for each speaker before the production take.

capture

Record in a quiet room with the same microphone position for the full track.
Capture a WAV master at 48 kHz or better before exporting web audio.
Slate the track ID and take number at the start of each raw file.
Record at least two clean takes for long dialogue tracks.

edit and export

Choose the clearest take; do not splice in a way that changes rhythm or vowel quality.
Remove obvious noise while preserving consonant onsets and final vowels.
Normalize loudness consistently across all course tracks.
Export replacement web audio to the same public path used by the current draft track.

release gate

Listen through once without the transcript and once while following the transcript.
Verify every line is present, in order, and spoken by the intended voice.
Check that lesson audio players, transcript links, and comprehension questions still match.
Commit the replacement audio, source notes, and generated site updates together.

acceptance

quality gates

gate

pronunciation

Vowels remain stable across slow and connected speech.
Kai spelling can be recovered from the audio without guessing English spelling habits.
Sacred Kai samples remain distinct from plain Common Kai pacing.

gate

pedagogy

The recording supports dictation, shadowing, and transcript-check practice.
Beginner audio is slower and clearer than advanced dialogue audio.
Course comprehension questions are still answerable from the recording.

gate

technical

No clipping, hum, abrupt cuts, or large volume jumps remain.
File names and public URLs remain stable for existing pages.
A raw master and exported web file are retained for every replaced track.