โ† All examples

ModelEthio-ASR Amharic (600M)1Ethio-ASR Multi (300M)5Ethio-ASR Multi (600M)4Ethio-ASR Multi (94M)6Meta MMS-1B7OmniASR-CTC (1B)2OmniASR-CTC (300M)9OmniASR-CTC (3B)4OmniASR-LLM (1B)5OmniASR-LLM (300M)6OmniASR-LLM (3B)3OpenAI Whisper-small7
Audio qualityClean Studio2Ambient Room Noise2Harsh Environment1Distant / Low-Volume Mic2Wideband Phone (VoLTE)4Randomized Phone Channel3Narrowband 2G + Packet Loss7

What did the model get wrong?

Ethio-ASR Multi (600M) Wideband Phone (VoLTE)   Example 10 of 20  ยท  speaker: Male

๐ŸŽง Listen โ€” Wideband Phone (VoLTE)
Compressed, narrow-band audio like a regular phone call.
โœ… What was actually said
แˆตแˆˆ แŒฅแ‰…แˆŽแ‰น แ‹แˆญแ‹แˆญ แˆ˜แˆจแŒƒ แ‹จแ‰ต แŠแ‹ แ‹จแˆ›แŒˆแŠ˜แ‹
๐Ÿค– What this model heard
amh แˆตแˆˆแŒฅแ‰…แˆแ‰ฑ แ‹แˆญแ‹แˆญ แŒƒแ‹จแ‰ต แŠแ‹ แ‹จแˆ›แŒˆแŠ˜แ‹
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
3
Correct
3
Wrong
1
Missed
0
Extra
43%
Words right
Out of 7 spoken words, this model got 3 right (43%). Errors: 3 words wrong; 1 missed.

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