โ† All examples

ModelEthio-ASR Amharic (600M)3Ethio-ASR Multi (300M)10Ethio-ASR Multi (600M)6Ethio-ASR Multi (94M)9Meta MMS-1B9OmniASR-CTC (1B)7OmniASR-CTC (300M)6OmniASR-CTC (3B)3OmniASR-LLM (1B)1OmniASR-LLM (300M)3OmniASR-LLM (3B)2OpenAI Whisper-small12
Audio qualityClean Studio43Ambient Room Noise18Harsh Environment9Distant / Low-Volume Mic26Wideband Phone (VoLTE)9Randomized Phone Channel22Narrowband 2G + Packet Loss12

What did the model get wrong?

OpenAI Whisper-small Narrowband 2G + Packet Loss   Example 8 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‰ แŒฃแˆ แŒฅแˆฉ แŠจแแ‰ฐแŠ›แ‹ แ‹จแแŒฅแŠแ‰ต แˆ˜แŒ แŠ• แŠฅแˆตแŠจ แˆตแŠ•แ‰ต แŠแ‹ แ‹จแˆšแ‹ฐแˆญแˆฐแ‹
๐Ÿค– What this model heard
bod emdurus khaqsit el mewfik enet mead emdans keshint en yw erse
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
0
Correct
9
Wrong
0
Missed
3
Extra
0%
Words right
Out of 9 spoken words, this model got 0 right (0%). Errors: 9 words wrong; 3 extra.

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