engineering/uncloseai-speech
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
uncloseai-speech/cloned-voices
History
russell@unturf.com c5247cd993
whisper_refs: normalize transcripts (sentence-case, strip quote artifacts) 
normalize_text() handles three Whisper quirks that produced messy F5
ref_texts:

  1. Lowercase output with no terminal period — capitalize first letter,
     append "." if missing.
  2. Hallucinated "' clusters Whisper inserts when it interprets a
     fragment as quoted dialogue (cora, ivan, atlas, hope had these).
     Strip everywhere; never legitimate English punctuation.
  3. Trailing apostrophe-then-period (.'.) from earlier rounds where a
     closing-quoted line got an extra "." appended — collapse to single
     terminal. Function is now idempotent.

Adds --from-cache flag: skip ASR, re-apply normalize from cached
whisper_refs.json. No GPU needed, useful after tuning the normalizer.
Lazy-imports torch so --from-cache works on any host.

Affects 9 of 40 voices: clara, grace, hazel, iris, felix, hugo
(lowercase fix); cora, ivan, atlas, hope (quote-cluster fix).
2026-05-24 16:02:18 -04:00
..
amber.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
archer.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
aria.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
atlas.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
blake.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
brooke.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
caleb.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
clara.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
cole.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
cora.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
dane.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
diana.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
eden.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
elena.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
ezra.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
faye.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
felix.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
finn.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
gemma.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
grace.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
grant.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
hazel.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
heath.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
hope.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
hugo.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
iris.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
ivan.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
ivy.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
jasper.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
jude.wav Expand to all 40 test-clean voices with idempotent registry 2026-01-27 15:30:42 -05:00
kai.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
leo.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
luna.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
marcus.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
maya.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
owen.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
ruby.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
sage.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
sofia.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
theo.wav Regenerate voices with upstream-verified genders from SPEAKERS.TXT 2026-01-27 13:22:26 -05:00
voices_metadata.json whisper_refs: normalize transcripts (sentence-case, strip quote artifacts) 2026-05-24 16:02:18 -04:00
whisper_refs.json whisper_refs: normalize transcripts (sentence-case, strip quote artifacts) 2026-05-24 16:02:18 -04:00