โ† All examples

ModelEthio-ASR Amharic (600M)9Ethio-ASR Multi (300M)13Ethio-ASR Multi (600M)10Ethio-ASR Multi (94M)9Meta MMS-1B12OmniASR-CTC (1B)6OmniASR-CTC (300M)8OmniASR-CTC (3B)4OmniASR-LLM (1B)1OmniASR-LLM (300M)9OmniASR-LLM (3B)4OpenAI Whisper-small12
Audio qualityClean Studio8Ambient Room Noise8Harsh Environment10Distant / Low-Volume Mic8Wideband Phone (VoLTE)9Randomized Phone Channel9Narrowband 2G + Packet Loss13

What did the model get wrong?

Ethio-ASR Multi (300M) Narrowband 2G + Packet Loss   Example 9 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‰แŒฅแˆฉ แ‹œแˆฎ แ‹˜แŒ แŠ แŠ แŠ•แ‹ต แˆแˆˆแ‰ต แŠ แˆแˆตแ‰ต แˆตแ‹ตแˆตแ‰ต แˆฐแ‰ฃแ‰ต แˆตแˆแŠ•แ‰ต แ‹˜แŒ แŠ แ‹œแˆฎ แŠแ‹
๐Ÿค– What this model heard
sid karo daro betani ani mila hanna hadi sawaa hana hate higaera ai
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
0
Correct
12
Wrong
0
Missed
1
Extra
0%
Words right
Out of 12 spoken words, this model got 0 right (0%). Errors: 12 words wrong; 1 extra.

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