Common Voice Scripted Speech 26.0 - British English

Splits and speakers: 215,340 validated clips total — train 201,330 clips / 3,244 speakers, dev 5,302 clips / 135 speakers, test 8,708 clips / 164 speakers. Split is speaker-disjoint (no client_id appears in more than one split), proportioned ~93.5/2.5/4.0 by clip count via a greedy per-speaker balancing algorithm (speakers have uneven clip counts, so exact 95/2.5/2.5 isn't achievable while keeping speakers disjoint).

Clip duration: 293.45 hours total (1,056,434 seconds) across 215,340 clips. Per-split: train 274.89h, dev 7.20h, test 11.36h. Per-clip: mean 4.91s, median 4.79s, min 0.70s, max 22.68s, std 1.69s.

Clip duration histogram:

   0-1s |                                                         11 ( 0.0%)
   1-2s | #####                                                4,327 ( 2.0%)
   2-3s | #########################                           24,030 (11.2%)
   3-4s | ############################################        41,997 (19.5%)
   4-5s | ##################################################  47,290 (22.0%)
   5-6s | ##############################################      43,545 (20.2%)
   6-7s | ###############################                     29,156 (13.5%)
   7-8s | ################                                    15,207 ( 7.1%)
   8-9s | #######                                              6,559 ( 3.0%)
  9-10s | ###                                                  2,545 ( 1.2%)
 10-12s | #                                                      623 ( 0.3%)
 12-15s |                                                         36 ( 0.0%)
 15-20s |                                                         12 ( 0.0%)
 20-30s |                                                          2 ( 0.0%)
   30s+ |                                                          0 ( 0.0%)

Gender distribution (shipped subset, self-declared, raw dataset values): male_masculine 152,936; female_feminine 35,016; blank/unspecified 27,368; transgender 9; non-binary 7; do_not_wish_to_say 4. No gender filter was applied — this reflects the natural distribution of the accent-filtered population.

Age distribution (shipped subset, self-declared): thirties 57,900; twenties 54,158; fourties 36,132; blank/unspecified 23,592; fifties 15,313; sixties 13,438; teens 11,038; seventies 3,374; eighties 290; nineties 105. No age filter was applied.

Accent tags are multi-select: Common Voice's accents column is pipe-delimited, so a clip can carry more than one tag. 12,777 of the 215,340 shipped clips (5.9%) carry an additional non-UK tag alongside their matched UK tag (e.g. "United States English|England English", "England English|New Zealand English") — this reflects speakers who self-reported a mixed/dual accent background, not a filtering error. 173 distinct accent tags appear in the shipped subset (presence-based, i.e. a clip is counted once per tag it carries); full distribution:

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 matches a case-insensitive substring on england, scot, wales/welsh, or ireland/irish — verified against the real accent tag distribution to introduce no false positives (e.g. no unrelated "New South Wales"-style matches). No gender or 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 were extracted from the ~2.58M-clip, ~89GB source archive (not the full archive) via a single sequential streaming pass over the source tar.gz.