โ† All examples

ModelEthio-ASR Amharic (600M)3Ethio-ASR Multi (300M)9Ethio-ASR Multi (600M)4Ethio-ASR Multi (94M)6Meta MMS-1B10OmniASR-CTC (1B)4OmniASR-CTC (300M)3OmniASR-CTC (3B)3OmniASR-LLM (1B)4OmniASR-LLM (300M)7OmniASR-LLM (3B)2OpenAI Whisper-small10
Audio qualityClean Studio6Ambient Room Noise4Harsh Environment7Distant / Low-Volume Mic6Wideband Phone (VoLTE)6Randomized Phone Channel4Narrowband 2G + Packet Loss8

What did the model get wrong?

OmniASR-LLM (300M) Harsh Environment   Example 12 of 20  ยท  speaker: Female

๐ŸŽง Listen โ€” Harsh Environment
Very heavy background noise โ€” much harder than typical audio.
โœ… What was actually said
แŠ แˆแŠ• แ‹ซแˆˆแˆแ‰ต แŠ แ‹ณแˆ› แАแ‹ แ‹ˆแ‹ฐ แŒ…แ‰กแ‰ฒ แˆˆแˆตแˆซ แˆแˆ„แ‹ต แˆตแˆˆแˆ†แА แАแ‹
๐Ÿค– What this model heard
แŠ แˆแŠ• แ‹ซแˆˆแ‹ แ‰ณแ‹ณแˆ› แАแ‹ แ‹ˆแ‹ฐ แŒ…แ‰กแ‰ฒ แˆŠ แ‹ซแˆตแˆซ แˆŠแˆ‚แ‹ต แˆฒ แˆ†แА แАแ‹
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
5
Correct
5
Wrong
0
Missed
2
Extra
50%
Words right
Out of 10 spoken words, this model got 5 right (50%). Errors: 5 words wrong; 2 extra.

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