Common Voice Scripted Speech 26.0 - American English (Female)

Splits and speakers: 115,209 validated clips total — train 109,047 clips / 1,674 speakers, dev 3,274 clips / 45 speakers, test 2,888 clips / 76 speakers. Split is speaker-disjoint (no client_id appears in more than one split), proportioned ~94.7/2.8/2.5 by clip count via a greedy per-speaker balancing algorithm targeting 95/2.5/2.5 (speakers have uneven clip counts, so exact 95/2.5/2.5 isn't achievable while keeping speakers disjoint).

Clip duration: 165.45 hours total (595,627 seconds) across 115,209 clips. Per-split: train 157.40h, dev 4.00h, test 4.05h. Per-clip: mean 5.17s, median 5.06s, min 0.94s, max 29.76s, std 1.68s.

Clip duration histogram:

   0-1s |                                                                    1 ( 0.0%)
   1-2s | ###                                                            1,324 ( 1.1%)
   2-3s | ######################                                         9,524 ( 8.3%)
   3-4s | ###############################################               19,735 (17.1%)
   4-5s | ############################################################  25,031 (21.7%)
   5-6s | ###########################################################   24,944 (21.7%)
   6-7s | ###########################################                   18,107 (15.7%)
   7-8s | #######################                                        9,970 ( 8.7%)
   8-9s | ##########                                                     4,488 ( 3.9%)
  9-10s | ####                                                           1,678 ( 1.5%)
 10-12s |                                                                  383 ( 0.3%)
 12-15s |                                                                   15 ( 0.0%)
 15-20s |                                                                    5 ( 0.0%)
 20-30s |                                                                    4 ( 0.0%)
   30s+ |                                                                    0 ( 0.0%)

Gender distribution: 100% female_feminine (115,209 clips).

Age distribution (shipped subset, self-declared): twenties 33,198; sixties 20,197; fifties 16,784; thirties 16,469; teens 16,425; fourties 9,102; blank/unspecified 1,905; seventies 1,086; eighties 43.

Accent tag QA: unlike British English's open-form (multi-select-tolerant) subset, this dataset uses closed-form exact-match accent filtering. 100% of the 115,209 shipped clips have accents == "United States English" (single clean value, no co-occurring tags) — verified programmatically against the staged train/dev/test TSVs. No multi-tag distribution table or overlap-fraction disclosure is needed here.

Open-form vs. closed-form comparison (why closed-form was chosen): restricted to gender == "female_feminine" across the full validated corpus, open-form matching (substring "united states", case-insensitive) would have yielded 118,477 clips / 1,842 speakers / 170.65h, vs. this closed-form subset's 115,209 clips / 1,795 speakers / 165.45h. Closed-form drops 3,268 clips (2.8%) in exchange for guaranteeing every shipped clip's speaker did not also self-identify with a different, co-occurring accent tag.

Derivation methodology: derived from the Mozilla Common Voice Scripted Speech 26.0 English release (cv-corpus-26.0-2026-06-12, locale en). Filter: validated rows only (present in the source validated.tsv), whose self-declared accents field, trimmed, is exactly "United States English" (closed-form, case-sensitive, no co-occurring pipe-delimited tags), and whose gender field is exactly "female_feminine". No age filter applied. Speaker-disjoint train/dev/test split via a greedy per-speaker assignment balancing toward 95/2.5/2.5 targets. Only the referenced MP3 clips (115,209 of them) were extracted from the ~94.6GB source archive (not the full archive) via a single sequential streaming pass over the source tar.gz.