Dataset Viewer
Auto-converted to Parquet Duplicate
audio_id
string
audio
audio
required_emotion
string
duration_seconds
float64
sample_rate
int64
channels
int64
11ZVLUNEfdKHzvfglFHy
angry
21.36
16,000
1
1tsLO8vY4CwFIXmcG8Fn
angry
36.66
16,000
1
1wqNzZ0X8afaWmppNtJj
angry
27.48
16,000
1
3XAPLLhesqPeOXyUYh5i
angry
23.46
16,000
1
4mgNEuGucwyijV8bMprB
angry
29.94
16,000
1
5a9Is5t1nQwoSrQo6ng6
angry
25.933
16,000
1
5aRSMdkUU7wbHu3kLYNb
angry
25.5
16,000
1
63aa7LnlCmvkVTMvAjfx
angry
25.38
16,000
1
8COq01h2tu3wgd014zYn
angry
18.18
16,000
1
8hb42shUo1axVdf6UolP
angry
23.82
16,000
1
8SaF1wfYeZTH54Wee5e2
angry
21
16,000
1
9nZS8pd84wV4PPXYpXZ0
angry
22.8
16,000
1
9POEQPkwpPDA1BW9y7fN
angry
20.94
16,000
1
AjulHMOjpIkvHQuVn7Sc
angry
27.06
16,000
1
APqVfnviTtUa3025FJrS
angry
32.233
16,000
1
AuFwuZZEeqcpdBQ0Fc9q
angry
20.04
16,000
1
EWa90Vl67THYQVri9jpM
angry
26.22
16,000
1
FRft4CK3WlOv3ZjBAynk
angry
26.28
16,000
1
GpBSVLLRVSqEU9bx6a3N
angry
35.1
16,000
1
GWF5g8Y3uoh3JYTtqNPK
angry
29.46
16,000
1
HiniQeL4A8MslPceoeQz
angry
31.62
16,000
1
HOVZ99Py5ip2z8ns3K38
angry
30.06
16,000
1
Hrg7s9HSTex24gCt8r8P
angry
18.78
16,000
1
Huj5jH4QPZlsnICM15Dy
angry
25.5
16,000
1
HurKiKdwiT3Oq5KilmTZ
angry
24.933
16,000
1
Iqxvlhc60uiMWvBfYK0s
angry
38.82
16,000
1
J8aatndkxVXalFgFIBiZ
angry
27.36
16,000
1
JhV8LuHsWiiwrXN3pIFz
angry
18.6
16,000
1
KcKP8ATmEwCbO8sMCjWG
angry
23.88
16,000
1
LmvX3zLIzpm0j33CfUwp
angry
25.62
16,000
1
LsBbS3bGKpF5agbinAn9
angry
28.08
16,000
1
MaLD2p9IuGHwMBAP9ZzG
angry
26.4
16,000
1
MLoGmNuHlJfTyLIco3le
angry
21.48
16,000
1
Mq2yURbwGdhJ3tGBqKfu
angry
22.5
16,000
1
Ov2Lnpb8e3JkFoFFnRS9
angry
18.84
16,000
1
OyGuCBe2gOdvkHwcyqCX
angry
28.56
16,000
1
Pe3uDXtHWShYyKZrf8dQ
angry
18.96
16,000
1
Pk0E1VeNwPUz4nbkjXdM
angry
31.44
16,000
1
QSOhMjPpXgn7BfMpEHQj
angry
23.76
16,000
1
RU6lWUxoLmUWFZrux7Qb
angry
22.02
16,000
1
0tMCzox5MTeqmmQkEhpI
disgusted
28.02
16,000
1
0YSK4ukzrgrkuCPO5lLM
disgusted
29.58
16,000
1
1xhRs9byaSrELHL7EfOe
disgusted
28.2
16,000
1
5mJS9x04GNd7dyr0GNKR
disgusted
30.36
16,000
1
7kss15f3LP8XF6E0h24Y
disgusted
31.14
16,000
1
8TT1fTxpQv23Hm6RKjDK
disgusted
26.52
16,000
1
9uKzL1xjprweMnPpsT1E
disgusted
30.9
16,000
1
aSlcNOqeF40kxA7ldQ2p
disgusted
24.24
16,000
1
ayRjbuy5Tc8zxjyPlRCL
disgusted
34.26
16,000
1
BkAaHO1lpqJGjmqChUMP
disgusted
32.4
16,000
1
buIlJEkipW5qxHyaJwjS
disgusted
21.66
16,000
1
cFYaMcqDH835n8N6qcI7
disgusted
32.64
16,000
1
ChOyXLzY6aA9YN98JnkZ
disgusted
36.3
16,000
1
CTC2rctUG4PHlDwmC8YN
disgusted
22.98
16,000
1
DD3xgnKT7knqRJg5Ptct
disgusted
38.58
16,000
1
FdKquYzV3JQj7I98k8pZ
disgusted
20.34
16,000
1
FNwe6m4KN2Luy4cPlBt8
disgusted
30.24
16,000
1
fPjWaYZJeJQsAhQJyU5V
disgusted
26.314
16,000
1
FWX5mf7BMkQL6QDWpdC1
disgusted
42.194
16,000
1
GhLqS3nXyCNBPzt5cY7Q
disgusted
26.58
16,000
1
HHgwsLkO39t8uxmgO67O
disgusted
37.2
16,000
1
HLoEknilueRMjUiS8VOu
disgusted
21.06
16,000
1
jjH5dfVQll6XXszDvLVr
disgusted
22.68
16,000
1
lOHa23nwVRYpDzjQXevi
disgusted
22.02
16,000
1
M8NgBYc1dCh5ASbdHc3B
disgusted
26.64
16,000
1
mEQine9suJOcLdEsAMwM
disgusted
19.74
16,000
1
NucGiNYxT59bNzbn6VvS
disgusted
26.46
16,000
1
p6kNUI6gdRdTe6P28DnJ
disgusted
21.54
16,000
1
rQxaEveJxVJTMJbv9yow
disgusted
34.44
16,000
1
Szc4lAhfFfa8JUDBI0Rd
disgusted
27.06
16,000
1
TbuLbD9Vy0AnkkQc8Nus
disgusted
28.5
16,000
1
Ua1WsdsHzgOUqhvbNigh
disgusted
24
16,000
1
UfJ7bROuEsxXAmkCMF4J
disgusted
31.974
16,000
1
VCvedsBDDYvfxgds6Xoo
disgusted
35.574
16,000
1
VGjsx5C5vE9qHvUATVJJ
disgusted
23.22
16,000
1
xJZHdSiT367tLIQ4Yr6f
disgusted
21.12
16,000
1
yL8AD43haj8EcZ1OL5ao
disgusted
40.86
16,000
1
YWDyJX7TsnGms6IRpVwT
disgusted
21.96
16,000
1
zFP5lOkP2Gm4Ndnkkva5
disgusted
25.02
16,000
1
ZIx0bsOXht3Cdo3UCpZ3
disgusted
33.714
16,000
1
2y3G3srKaAz4YUu3gnHj
fearful
31.62
16,000
1
4EjRUutkQizb0O4nTIu6
fearful
30.654
16,000
1
5pEbAaNf2DAJMR5RIWHN
fearful
28.654
16,000
1
5SCJEU64qorWgnFilL1a
fearful
25.56
16,000
1
9fJQ2HMNj8N5wZlrZpCT
fearful
18.48
16,000
1
9JYPITYzW6dZU2Itj7SO
fearful
39.12
16,000
1
9wSxUc3PbJsT2MaHLiQW
fearful
25.2
16,000
1
ahy5OGTRjSsDInyh3X4s
fearful
29.654
16,000
1
CYtKlWESrZRxddBxNCtf
fearful
28.634
16,000
1
DBIkc9DHsPf8AKSbD9aA
fearful
31.2
16,000
1
dlTmBKElR8gaJL159ddo
fearful
26.94
16,000
1
DwfNM4DNmu8UltgDBiZo
fearful
22.92
16,000
1
EJndPEIxUshUSU5FSvjO
fearful
23.1
16,000
1
ewZKTzqvrmhuPUTDOPFs
fearful
20.04
16,000
1
exOUefvjCsdfs8AuU0P3
fearful
32.34
16,000
1
FRI7S8lAljeQRhPgZHRI
fearful
19.68
16,000
1
FVnlfNMwgUCQRoHJ4Ktk
fearful
22.32
16,000
1
g3cQL72TawaRRvmpKN8G
fearful
20.64
16,000
1
GjWCBOIMonS8rwTBnX7c
fearful
32.94
16,000
1
H1mltpDz8zoJS2JXdUZr
fearful
30.48
16,000
1
End of preview. Expand in Data Studio

VocalAffectBench

VocalAffectBench is a test-only benchmark for evaluating whether AI audio models can identify expressed vocal emotion from raw audio.

Paper: VocalAffectBench: Evaluating Vocal Emotion Recognition in AI Audio Models

The benchmark targets the expressed emotion — what the speaker conveys through vocal tone, prosody, pace, intensity, and pauses — not inferred internal state.

Contents

  • 280 human-recorded English WAV clips, totalling 2.32 hours.
  • 7 emotion classes, balanced at 40 clips each: angry, disgusted, fearful, happy, neutral, sad, surprised.
  • 6 tracked baseline model outputs.
  • Audio recorded at 16 kHz mono.

Released files:

  • audio/*.wav — benchmark audio files, named by audio_id.
  • data/metadata.jsonl — per-clip metadata: emotion label, duration, sample rate.
  • data/audio-metadata.csv — flat CSV version of clip metadata.
  • data/predictions.csv — all baseline model predictions.
  • data/leaderboard-summary.csv — aggregate baseline leaderboard.
  • baselines/predictions/*.csv — per-model prediction files.
  • baselines/results.csv — aggregate leaderboard (same as above).
  • paper/ — paper PDF and LaTeX source.

The public release does not include transcripts, script text, speaker identities, or demographic metadata.

Task

Each item is an audio clip of a person expressing one of the seven target emotions. Models are evaluated under a raw-audio-only protocol — the model receives only the audio file and a fixed prompt listing the allowed labels. No transcripts, speaker metadata, domain hints, or post-processing are permitted in the main setting.

The prompt instructs the model to:

Choose exactly one primary expressed emotion from the allowed label set. Base your answer only on the expressed vocal tone, prosody, pace, intensity, pauses, and wording. Do not infer the speaker's private internal state.

Label Set

angry  |  disgusted  |  fearful  |  happy  |  neutral  |  sad  |  surprised

Metrics

VocalAffectBench reports overall accuracy (correct / total scored clips). Because the benchmark is balanced at 40 clips per class, overall accuracy equals macro accuracy — no separate macro metric is needed.

Baselines

Current tracked baselines (6 models):

Rank Model Overall Accuracy
1 gemini_3_5_flash 44.3%
2 hume_prosody 38.0%
3 tr_qwen3_5_omni_plus 37.9%
4 tr_voxtral_small 34.3%
5 inworld_voice_profile 28.6%
6 openai_realtime 27.9%

All baselines use the raw-audio-only protocol with no transcripts or post-processing.

Data Format

metadata.jsonl

Each line is a JSON object:

{
  "audio_id": "06Eg9dXO99fAAti4HB34",
  "file_name": "audio/06Eg9dXO99fAAti4HB34.wav",
  "required_emotion": "neutral",
  "duration_seconds": 12.4,
  "sample_rate": 16000,
  "channels": 1
}

predictions.csv

audio_id, required_emotion, model_name, provider_model, predicted_label, mapped_label, confidence, correct, error
  • predicted_label — raw label returned by the model.
  • mapped_label — normalised to the 7-class label set.
  • correcttrue / false.

Use and Limits

VocalAffectBench is intended for diagnostic evaluation, provider comparison, and regression tracking of audio emotion models. It is not intended as a training corpus, hidden leaderboard, universal emotion-quality measure, or biometric dataset.

Emotion labels reflect the expressed vocal performance, not speaker demographics or inferred mental states.

Citation

@misc{vocalaffectbench2026,
  title  = {VocalAffectBench: Evaluating Vocal Emotion Recognition in AI Audio Models},
  author = {Debaupte, Luc and Baumgartner, Tyler and Tai, Brandon and Fan, Candice and Wang, Bill and Zhong, Yi},
  year   = {2026},
  note   = {Benchmark dataset}
}

License

VocalAffectBench is released under the MIT License.

Downloads last month
387