Dataset
Lwazi
ANV
← All examples
Model summary →
Model
facebook/omniASR-CTC-1B
Open
12
facebook/omniASR-CTC-300M
Open
14
facebook/omniASR-CTC-3B
Open
15
facebook/omniASR-CTC-7B
Open
13
facebook/omniASR-LLM-1B
Open
11
facebook/omniASR-LLM-300M
Open
14
facebook/omniASR-LLM-3B
Open
13
facebook/omniASR-LLM-7B
Open
11
aadel4/omniASR-CTC-1B-v2
Open
14
aadel4/omniASR-CTC-300M-v2
Open
17
badrex/w2v-bert-2.0-zulu-asr
Open
14
dsfsi/anv-whisper-large-v3-turbo-anv-zul
Open
14
dsfsi/anv-whisper-small-anv-zulu-first-batch
Open
21
facebook/mms-1b-all
Open
17
openai/whisper-large-v3
Open
20
openai/whisper-small
Open
148
sitwala/whisper-large-anv-sot
Open
22
sitwala/whisper-large-v3-anv-sot
Open
27
sitwala/whisper-large-v3-turbo-anv-zul-150h
Open
17
sitwala/whisper-large-v3-turbo-anv-zul-250h
Open
14
sitwala/whisper-large-v3-turbo-anv-zul-50h
Open
16
What did the model get wrong?
sitwala/whisper-large-anv-sot
Open
Example 46 of 100 · sample index 45
✅ What was actually said (isiZulu reference)
zonke
izinsuku,
[s]
ubani
engumna-
[n],
ee
angiboni
mina
[n]
engumNaziri
oh,
zonke
izinsuku
engumNaziri,
ka
Jehova,
akayi
kusondela
esidunjini
[s].
🤖 What this model heard
zonke
e
zentsoko
e
ngomna
e
ngomna
zere
o
zonke
e
zentsoko
e
ngomna
zere
ka
tshehova
a
ka
e
kosondela
e
setong
jeng
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
2
Correct
19
Wrong
0
Missed
3
Extra
10%
Words right
1.158
WER
0.496
CER
18.40s
Duration
0.1s
Latency