โ† All examples

ModelEthio-ASR Amharic (600M)4Ethio-ASR Multi (300M)15Ethio-ASR Multi (600M)11Ethio-ASR Multi (94M)14Meta MMS-1B15OmniASR-CTC (1B)5OmniASR-CTC (300M)9OmniASR-CTC (3B)3OmniASR-LLM (1B)4OmniASR-LLM (300M)5OmniASR-LLM (3B)12OpenAI Whisper-small14
Audio qualityClean Studio4Ambient Room Noise4Harsh Environment6Distant / Low-Volume Mic4Wideband Phone (VoLTE)5Randomized Phone Channel6Narrowband 2G + Packet Loss14

What did the model get wrong?

Ethio-ASR Multi (94M) Narrowband 2G + Packet Loss   Example 4 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
แŠฅแˆฝ แŒแˆญแˆ› แŠ แŒˆแˆแŒแˆŽแ‰ฑแŠ• แˆˆแˆ›แ‰‹แˆจแŒฅ แˆตแ‰ถแ• แ‰ข แ‰ฅแˆˆแ‹ แ‹ˆแ‹ฐ แˆตแ‹ตแˆตแ‰ต แˆ˜แ‰ถ แŠ แˆตแˆซ แŠ แˆแˆตแ‰ต แŠ แŒญแˆญ แ‰แŒฅแˆญ แ‹ญแˆ‹แŠฉ
๐Ÿค– What this model heard
amh แŠฅแˆฝ แ‰ แˆญแˆ› แŠ แŒˆแˆแŒแˆŽแ‰ตแŠ• แˆˆแˆ˜แŒ“แˆจแ‰ต แˆตแ‰ณแ‰ตแŠ แ‰ฅแˆˆแ‹ แ‹ฐ แˆตแ‹ตแˆต แˆ˜แ‰… แŠ แŠ แŠ•แ‹ต แŠ แ‰€แ‹ญ แ‰ แˆญ แˆแŠ“แแˆ‰
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
1
Correct
13
Wrong
1
Missed
0
Extra
7%
Words right
Out of 15 spoken words, this model got 1 right (7%). Errors: 13 words wrong; 1 missed.

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