Overview

Corpus-wide manifest balance statistics

Language Completion

Script Type Mix

Quality Checks

Balance Scorecard

Score = 100 × (1 − coefficient of variation). ≥85 balanced, ≥60 moderate.

Parameter Score Status Categories Min – Max

Target vs Actual (S1–S4)

Per-Language Script Types

Domain Distribution

Emotion Distribution

Content Type

Gender Balance

Voice Types

Direction Slot Scores

Regions (top 12)

Top Assigned Audio Events

Top Filler Words

S4 Phoneme Targets

Emotions

# Emotion Count Status

Audio Event Tags

# Marker Tag Category Count Status

Recording Instructions

Per-record direction text for the recording artist (2–4 sentences). Nearly unique per script.

# Instruction Count Sample ID

Voice Description — One Line

Short persona voice summary (`voice_description.one_line`).

# One Line Count Sample ID

Voice Description — Full

Extended persona voice profile (`voice_description.full`).

# Full Description Count Sample ID

Direction Prompt Text

Recording direction prompt (`direction.prompt_text`) — delivery, emotion, and pacing cues.

# Prompt Text Count Sample ID

All Batches

Batch Records Target Completion Domain Score Emotion Score Failures Updated

Corpus Growth