โ† All examples

ModelEthio-ASR Amharic (600M)7Ethio-ASR Multi (300M)15Ethio-ASR Multi (600M)6Ethio-ASR Multi (94M)10Meta MMS-1B12OmniASR-CTC (1B)12OmniASR-CTC (300M)7OmniASR-CTC (3B)12OmniASR-LLM (1B)7OmniASR-LLM (300M)7OmniASR-LLM (3B)8OpenAI Whisper-small12
Audio qualityClean Studio12Ambient Room Noise11Harsh Environment12Distant / Low-Volume Mic12Wideband Phone (VoLTE)12Randomized Phone Channel10Narrowband 2G + Packet Loss12

What did the model get wrong?

OpenAI Whisper-small Narrowband 2G + Packet Loss   Example 18 of 20  ยท  speaker: Female

๐ŸŽง Listen โ€” Narrowband 2G + Packet Loss
Heavy background noise plus phone-call compression โ€” worst-case stress test.
โœ… What was actually said
แ‰แŒฅแˆฌ แ‹œแˆฎ แ‹˜แŒ แŠ แŠ แˆแˆตแ‰ต แŠ แŠ•แ‹ต แŠ แŠ•แ‹ต แˆตแˆแŠ•แ‰ต แŠ แˆแˆตแ‰ต แŠ แŠ•แ‹ต แŠ แˆซแ‰ต แŠ แŠ•แ‹ต แŠแ‹
๐Ÿค– What this model heard
sqrl rurwetan amlum anq anq sqrlum amlum anq aran anq dillaw
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
0
Correct
11
Wrong
1
Missed
0
Extra
0%
Words right
Out of 12 spoken words, this model got 0 right (0%). Errors: 11 words wrong; 1 missed.

Try switching Audio quality โ€” see the same model degrade as the audio gets noisier.