System Detail
speechbrain/xvector
xvector
Submitted by: Public checkpoint
- Global position
- #21 / 23
- Scenario-macro EER
- 23.3539
- Scenario-macro minDCF
- 0.776078
- Trial-set coverage
- 26 / 26
- Scenario top-three
- 0
Publicly ranked
This result is included in the reviewed public ranking.
Model Provenance
Source and architecture
- Displayed name:
speechbrain/xvector - Submitted by: Public checkpoint
- Metadata: OpenSVBench model record
- Source: SpeechBrain
- Architecture: x-Vector TDNN
- Model family: x-Vector
- Submitted alias:
spkrec-xvect-voxceleb - Created at:
2026-05-29T14:42:47.406285+00:00
Training And Links
Training data and references
- Training data: VoxCeleb1 + VoxCeleb2 training data
- Training setup: SpeechBrain TDNN x-vector release with statistical pooling and categorical cross-entropy training.
- Submission mode:
full-core - Model source: Open link
- Paper: X-VECTORS: ROBUST DNN EMBEDDINGS FOR SPEAKER RECOGNITION
Training/evaluation overlap
This checkpoint was trained or fine-tuned on VoxCeleb1. Results on VoxCeleb1-E and VoxCeleb1-H are not independent of the training data.
Higher-Ranked Scenarios
Scenario ranks above the global position
Show 3 supporting trial sets
- VoxPopuli Aging: #18 on its trial-set ranking.
- Lombard Grid Lombard: #20 on its trial-set ranking.
- VOiCES Noise/Reverb: #20 on its trial-set ranking.
Lower-Ranked Scenarios
Scenario ranks below the global position
Show 3 supporting trial sets
- 3D-Speaker Device: #23 on its trial-set ranking.
- 3D-Speaker Dialect: #23 on its trial-set ranking.
- 3D-Speaker Distance: #23 on its trial-set ranking.
Scenario Rankings
Full ranking across scenarios
This table shows where the model sits on each scenario, using source-balanced scenario scores and the linked trial sets as evidence.
| Scenario | Rank | Lens | Evidence | Score |
|---|---|---|---|---|
|
Aging Robustness
How stable identity representations remain across age-derived and longitudinal recording gaps.
|
#20/23
Needs work
|
Speaker aging and time gaps
13.6725 EER from leader
|
15.4642
0.661325 minDCF
|
|
|
Noise / Reverb Robustness
How reliably the model preserves identity when room acoustics, distractor noise, and microphone placement deviate from an easier in-corpus reference condition.
|
#20/23
Needs work
|
Noise and reverberation
16.8200 EER from leader
|
19.0800
0.568120 minDCF
|
|
|
Cross-Lingual Robustness
How well speaker identity survives enrollment-test language mismatch.
|
#21/23
Needs work
|
Language mismatch
12.4412 EER from leader
|
14.0989
0.582957 minDCF
|
|
|
Speaking-Style Robustness
Whether the model can preserve identity across emotion-driven change, whispered speech, and noise-induced Lombard speaking style.
|
#21/23
Needs work
|
Speaking style shift
13.7515 EER from leader
|
15.6681
0.597160 minDCF
|
|
|
Overlap Robustness
How well speaker identity survives light, mid, and heavy overlap in both meeting and domestic recordings.
|
#21/23
Needs work
|
Overlapping speakers
15.1271 EER from leader
|
23.2363
0.916343 minDCF
|
|
|
Distance Robustness
How much performance changes across meeting and domestic distance mismatch.
|
#21/23
Needs work
|
Distance mismatch
19.3893 EER from leader
|
24.6461
0.923625 minDCF
|
|
|
Short-Duration Robustness
How much performance holds up when speech evidence is limited by duration.
|
#21/23
Needs work
|
Short-duration speech
16.0442 EER from leader
|
25.0788
0.830830 minDCF
|
|
|
Channel / Device Robustness
How resilient the model is to device and channel mismatch.
|
#21/23
Needs work
|
Device and channel variation
27.0203 EER from leader
|
31.0227
0.964547 minDCF
|
|
|
Genre-Shift Robustness
Whether performance holds when CN-Celeb enrollment and test speech come from different source genres.
|
#21/23
Needs work
|
Source-genre variation
22.5132 EER from leader
|
36.7857
0.987563 minDCF
|
|
|
In-The-Wild Robustness
How strong the model is on unconstrained celebrity and media speech across official CN-Celeb and VoxCeleb protocols.
|
#22/23
Needs work
|
Open-domain media speech
13.7466 EER from leader
|
17.4683
0.675055 minDCF
|
|
|
Accent / Dialect Robustness
How reliably the model tracks identity across accent and dialect mismatch.
|
#23/23
Needs work
|
Accent and dialect variation
17.5220 EER from leader
|
34.3436
0.829337 minDCF
|
Trial-Set Results
Raw ranking by trial set
Expand this section to inspect the detailed trial-set rankings.
Show
Raw ranking by trial set
Expand this section to inspect the detailed trial-set rankings.
| Trial Set | Rank | Scenarios | Trials | Score |
|---|---|---|---|---|
|
TidyVoiceX2-ASV
TidyVoiceX2-ASV
|
#21/23
|
Cross-lingual
|
200000 |
14.0989
0.582957 minDCF
|
|
HI-MIA
HI-MIA
|
#21/23
|
Short-duration
|
660000 |
12.0075
0.621905 minDCF
|
|
GSC Short
Google Speech Commands
|
#21/23
|
Short-duration
|
220000 |
17.5650
0.848580 minDCF
|
|
GLOBE
GLOBE
|
#21/23
|
Accent/dialect
|
56848 |
27.2639
0.658882 minDCF
|
|
CN-Celeb
CN-Celeb
|
#21/23
|
In-the-wild
|
3484292 |
24.2749
0.826837 minDCF
|
|
CN-Celeb Genre
CN-Celeb
|
#21/23
|
Genre shift
|
440000 |
36.7857
0.987563 minDCF
|
|
CN-Celeb Short
CN-Celeb
|
#23/23
|
Short-duration
|
546964 |
37.8590
0.985247 minDCF
|
|
VoxCeleb1-O
VoxCeleb1
|
#22/23
|
In-the-wild
|
37611 |
8.9618
0.489966 minDCF
|
|
VoxCeleb1-E
VoxCeleb1
|
#22/23
|
In-the-wild
|
579818 |
9.1649
0.478791 minDCF
|
|
VoxCeleb1-H
VoxCeleb1
|
#22/23
|
In-the-wild
|
550894 |
13.8588
0.601061 minDCF
|
|
VoxCeleb Short
VoxCeleb1
|
#23/23
|
Short-duration
|
394724 |
32.8837
0.867590 minDCF
|
|
3D-Speaker Device
3D-Speaker
|
#23/23
|
Channel/device
|
180000 |
42.6314
0.999900 minDCF
|
|
3D-Speaker Distance
3D-Speaker
|
#23/23
|
Distance
|
175163 |
39.6313
1.000000 minDCF
|
|
3D-Speaker Dialect
3D-Speaker
|
#23/23
|
Accent/dialect
|
180000 |
41.4233
0.999793 minDCF
|
|
FFSVC 2022 Cross-Channel
FFSVC 2022
|
#21/23
|
Channel/device
|
72000 |
19.4139
0.929194 minDCF
|
|
FFSVC 2022 Cross-Domain
FFSVC 2022
|
#21/23
|
Distance
|
66546 |
19.5093
0.933937 minDCF
|
|
Whisper40 Whisper
Whisper40
|
#21/23
|
Speaking style
|
17600 |
32.1250
0.999063 minDCF
|
|
Lombard Grid Lombard
Lombard Grid
|
#20/23
|
Speaking style
|
29524 |
1.8405
0.099143 minDCF
|
|
VOiCES Noise/Reverb
VOiCES
|
#20/23
|
Noise/reverb
|
55000 |
19.0800
0.568120 minDCF
|
|
CHiME-6 Domestic Far-Field
CHiME-6
|
#20/23
|
Distance
|
36487 |
23.8288
0.972807 minDCF
|
|
CHiME-6 Overlap
CHiME-6
|
#22/23
|
Overlap
|
39600 |
30.1667
0.964667 minDCF
|
|
ESD
ESD
|
#22/23
|
Speaking style
|
437408 |
13.0389
0.693274 minDCF
|
|
AliMeeting Near/Far
AliMeeting
|
#21/23
|
Distance
|
220000 |
15.6150
0.787755 minDCF
|
|
AliMeeting Overlap
AliMeeting
|
#21/23
|
Overlap
|
165000 |
16.3060
0.868020 minDCF
|
|
VoxKnesset
VoxKnesset
|
#21/23
|
Aging
|
158312 |
22.2605
0.834057 minDCF
|
|
VoxPopuli Aging
VoxPopuli
|
#18/23
|
Aging
|
146575 |
8.6679
0.488593 minDCF
|