← All examples

ModelEthio-ASR Amharic (600M)17Ethio-ASR Multi (300M)19Ethio-ASR Multi (600M)13Ethio-ASR Multi (94M)19Meta MMS-1B19OmniASR-CTC (1B)19OmniASR-CTC (300M)19OmniASR-CTC (3B)19OmniASR-LLM (1B)15OmniASR-LLM (300M)10OmniASR-LLM (3B)10OpenAI Whisper-small19
Audio qualityClean Studio15Ambient Room Noise2Harsh Environment0Distant / Low-Volume Mic14Wideband Phone (VoLTE)1Randomized Phone Channel0Narrowband 2G + Packet Loss19

What did the model get wrong?

OmniASR-CTC (300M) Narrowband 2G + Packet Loss   Example 1 of 20  Ā·  speaker: Female

šŸŽ§ Listen — Narrowband 2G + Packet Loss
Heavy background noise plus phone-call compression — worst-case stress test.
āœ… What was actually said
ሲ ሲ ቲ įˆįˆˆį‰µ į‹œįˆ® įˆįˆˆį‰µ įŠ įˆįˆµį‰µ įŠ įŠ•į‹µ įˆįˆˆį‰µ įŠ įŠ•į‹µ įˆ°į‰£į‰µ įˆįˆˆį‰µ įŠ įˆ«į‰µ įˆ°į‰£į‰µ įˆµįˆįŠ•į‰µ į‹œįˆ® įˆ°į‰£į‰µ į‹˜įŒ įŠ įŠ įˆ«į‰µ
šŸ¤– What this model heard
f f l vr l smi an l an sabt l saraት sባann vr sባt jtmaraት
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
0
Correct
16
Wrong
3
Missed
0
Extra
0%
Words right
Out of 19 spoken words, this model got 0 right (0%). Errors: 16 words wrong; 3 missed.

Try switching Audio quality — see the same model degrade as the audio gets noisier.