engineering/uncloseai-speech
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix voice gender assignments - verified against LibriSpeech SPEAKERS.TXT

Browse source
This commit is contained in:
russell@unturf.com 2026-01-27 13:13:07 -05:00
parent c56b8cb24e
commit b77af6d603
1 changed files with 23 additions and 23 deletions
Download patch file Download diff file Expand all files Collapse all files

46
scripts/download_diverse_voices.py
View file

@ -24,62 +24,62 @@ except ImportError:
print("Install required libraries: pip install datasets soundfile")
# LibriSpeech test-clean speaker metadata
# Curated for voice diversity - different genders, tones, pacing
# Gender verified against official SPEAKERS.TXT from OpenSLR
# Format: speaker_id: (gender, description)
SPEAKERS = {
# Female speakers (11)
"1089": ("female", "neutral, balanced - clear professional delivery"),
"1188": ("female", "warm, friendly - approachable delivery"),
# Female speakers (11) - verified from SPEAKERS.TXT
"1221": ("female", "expressive, storyteller - engaging narration"),
"1320": ("female", "soft, gentle - calm and soothing"),
"1284": ("female", "clear, measured - precise enunciation"),
"1580": ("female", "clear narrator - smooth reading style"),
"2094": ("female", "warm, rich - resonant tone"),
"3570": ("female", "bright, energetic - lively and upbeat"),
"3575": ("female", "calm, neutral - steady and even"),
"4446": ("female", "mature, authoritative - confident tone"),
"4507": ("female", "young, clear - crisp enunciation"),
"4970": ("female", "smooth, polished - announcer style"),
"5142": ("female", "warm, conversational - natural flow"),
"6829": ("female", "crisp, articulate - precise diction"),
"6930": ("female", "smooth, polished - radio-style delivery"),
# Male speakers (10)
"1284": ("male", "clear, measured - precise enunciation"),
"1580": ("male", "clear narrator - smooth reading style"),
"2094": ("male", "warm baritone - rich and resonant"),
# Male speakers (10) - verified from SPEAKERS.TXT
"1089": ("male", "neutral, balanced - clear professional delivery"),
"1188": ("male", "warm, friendly - approachable delivery"),
"1320": ("male", "soft, gentle - calm and soothing"),
"2830": ("male", "young, energetic - dynamic delivery"),
"4077": ("male", "mature, professor - thoughtful pacing"),
"4970": ("male", "smooth announcer - polished delivery"),
"5105": ("male", "neutral, clear - straightforward style"),
"5639": ("male", "deep, dramatic - authoritative tone"),
"6930": ("male", "smooth, polished - radio-style delivery"),
"7021": ("male", "light, conversational - casual tone"),
"7176": ("male", "warm storyteller - engaging narrative"),
}
# 21 voices: gendered names that match the speaker
# Female names for female speakers, male names for male speakers
# Gender verified against LibriSpeech SPEAKERS.TXT
VOICE_MAP = {
# Female voices (11)
"aria": "1089", # F - neutral, balanced
"clara": "1188", # F - warm, friendly
"elena": "1221", # F - expressive, storyteller
"grace": "1320", # F - soft, gentle
# Female voices (11) - all verified female speakers
"aria": "1221", # F - expressive, storyteller
"clara": "1284", # F - clear, measured
"elena": "1580", # F - clear narrator
"grace": "2094", # F - warm, rich
"hazel": "3575", # F - calm, neutral
"iris": "4507", # F - young, clear
"luna": "3570", # F - bright, energetic
"maya": "5142", # F - warm, conversational
"ruby": "4446", # F - mature, authoritative
"sage": "6829", # F - crisp, articulate
"sofia": "6930", # F - smooth, polished
"sofia": "4970", # F - smooth, polished
# Male voices (10)
# Male voices (10) - all verified male speakers
"atlas": "5639", # M - deep, dramatic
"caleb": "1284", # M - clear, measured
"caleb": "1089", # M - neutral, balanced
"felix": "7021", # M - light, conversational
"hugo": "7176", # M - warm storyteller
"jasper": "4077", # M - mature, professor
"kai": "2830", # M - young, energetic
"leo": "4970", # M - smooth announcer
"marcus": "2094", # M - warm baritone
"leo": "6930", # M - smooth, polished
"marcus": "1188", # M - warm, friendly
"owen": "5105", # M - neutral, clear
"theo": "1580", # M - clear narrator
"theo": "1320", # M - soft, gentle
}