Cantonese Voice Benchmark粵語語音基準
EN繁中

Cantonese Voice Benchmark 2026.1 · Qualification Round

Leaderboard

Every published number carries a bootstrap confidence interval. Entries that overlap are reported as tied rather than ordered — with 30 clips a fraction of a percent is noise.

Window
Qualification Round
Closes
30 Sept, 23:59 HKT
Entrants
10
Updated
24 Sept, 01:20 HKT
Ranking
Core ASR + LLM + TTSFull all pillars
Open the card
Core Score (ASR + LLM + TTS). Entries whose 95% confidence intervals overlap share a rank band and are reported as tied.
 Rank Team CoreASRCERLLMJudgeTTSMOSE2ENatural TTFB
1tied
MediConCenliveNamed
95.6
13.3%10.2–17.2
4.90 / 54.77–5.00
5.00 / 5measured, not rated
—Organisers run this
4.64sp95 6.30
1
Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Sampan10live
94.7
19.8%15.6–24.4
5.00 / 55.00–5.00
5.00 / 5measured, not rated
—Not entered
0.18sp95 0.76
1
Ener-GliveNamed
94.1
15.8%12.3–19.8
4.80 / 54.61–4.95
5.00 / 5measured, not rated
—Not entered
1.31sp95 2.71
4tied
Vibran-CliveNamed
91.6
19.5%16.4–23.3
5.00 / 55.00–5.00
4.68 / 5measured, not rated
—Not entered
0.41sp95 1.65
4
Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Osprey8live
90.9
19.4%16.4–22.8
4.53 / 54.14–4.89
5.00 / 5measured, not rated5.00–5.00
—Not entered
5.17sp95 6.67
4
DannySliveNamed
90.8
19.7%16.3–25.2
4.91 / 54.77–5.00
4.68 / 5measured, not rated
—Organisers run this
4.17sp95 5.64
4
Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Vanguard8live
89.9
26.7%23.2–30.7
4.86 / 54.67–5.00
4.82 / 5measured, not rated
—Not entered
0.81sp95 4.79
8
Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Petrel8live
79.4
52.3%48.4–56.5
4.59 / 54.20–4.88
4.68 / 5measured, not rated
—Organisers run this
16.48sp95 21.48
—
Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Yardarm9live
—ASR — scored, not submitted
—Scored, not submitted
4.89 / 54.73–5.00
—Not entered
—Not entered
—
—
Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Gannet8live
—ASR — scored, not submitted
—Scored, not submitted
—Scored, not submitted
—Scored, not submitted
—Organisers run this
—
Reference baselinesIndicative figures for scale. Not measured on this corpus — no entrant is ranked against them.
A
Open-source pipeline
76.2
11.1%
3.81 / 5
3.44 / 5
3.29 / 5
2.11sp95 3.88
B
Proprietary voice LLM
82.7
8.6%
4.14 / 5
3.62 / 5
4.01 / 5
0.88sp95 1.52
1MediConCenlive95.6
ASR
13.3%
10.2–17.2
LLM
4.90
4.77–5.00
TTS
5.00
measured, not rated
E2E
—
Organisers run this
TTFB 4.64s p50 · 6.30s p95
1Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Sampan10live94.7
ASR
19.8%
15.6–24.4
LLM
5.00
5.00–5.00
TTS
5.00
measured, not rated
E2E
—
Not entered
TTFB 0.18s p50 · 0.76s p95
1Ener-Glive94.1
ASR
15.8%
12.3–19.8
LLM
4.80
4.61–4.95
TTS
5.00
measured, not rated
E2E
—
Not entered
TTFB 1.31s p50 · 2.71s p95
4Vibran-Clive91.6
ASR
19.5%
16.4–23.3
LLM
5.00
5.00–5.00
TTS
4.68
measured, not rated
E2E
—
Not entered
TTFB 0.41s p50 · 1.65s p95
4Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Osprey8live90.9
ASR
19.4%
16.4–22.8
LLM
4.53
4.14–4.89
TTS
5.00
measured, not rated
5.00–5.00
E2E
—
Not entered
TTFB 5.17s p50 · 6.67s p95
4DannySlive90.8
ASR
19.7%
16.3–25.2
LLM
4.91
4.77–5.00
TTS
4.68
measured, not rated
E2E
—
Organisers run this
TTFB 4.17s p50 · 5.64s p95
4Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Vanguard8live89.9
ASR
26.7%
23.2–30.7
LLM
4.86
4.67–5.00
TTS
4.82
measured, not rated
E2E
—
Not entered
TTFB 0.81s p50 · 4.79s p95
8Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Petrel8live79.4
ASR
52.3%
48.4–56.5
LLM
4.59
4.20–4.88
TTS
4.68
measured, not rated
E2E
—
Organisers run this
TTFB 16.48s p50 · 21.48s p95
—Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Yardarm9live—
ASR — scored, not submitted
ASR
—
Scored, not submitted
LLM
4.89
4.73–5.00
TTS
—
Not entered
E2E
—
Not entered
—Anonymous by default. Being named is the entrant’s choice, reversible until results publish; after that a named entry can still be un-named on request.Gannet8live—
ASR — scored, not submitted
ASR
—
Scored, not submitted
LLM
—
Scored, not submitted
TTS
—
Scored, not submitted
E2E
—
Organisers run this
Reference baselinesIndicative figures for scale. Not measured on this corpus — no entrant is ranked against them.
AOpen-source pipeline76.2
BProprietary voice LLM82.7
How we measurePermanent link to this editionSnapshot 24 Sept, 01:20 HKT · scoring v1.2
Figure
ScoresIntervals
Core Score out of 100, higher is better. The same numbers as the table, without the uncertainty — switch to Intervals to see which entries this data cannot separate. 2 entries are not shown: they have no Core Score yet.
TiedOverlapping 95% confidence intervals. Reported as tied.
Not enteredDid not compete on that pillar. Never a zero.
PendingAwaiting human evaluation.
Insufficient dataUnder 80% of items completed without a platform fault. No score, no quota consumed.
Scored, not submittedThe team has a scored run it has not submitted. Only submitted runs are published; the latest one submitted is shown.
No runs yetEntered, not run yet. Conversation (E2E) is run by the organisers.
measured, not ratedMeasured by machine from the returned audio. No listening panel has rated it yet.