โ† All examples

ModelEthio-ASR Amharic (600M)1Ethio-ASR Multi (300M)7Ethio-ASR Multi (600M)3Ethio-ASR Multi (94M)7Meta MMS-1B9OmniASR-CTC (1B)2OmniASR-CTC (300M)7OmniASR-CTC (3B)2OmniASR-LLM (1B)2OmniASR-LLM (300M)6OmniASR-LLM (3B)3OpenAI Whisper-small9
Audio qualityClean Studio5Ambient Room Noise4Harsh Environment10Distant / Low-Volume Mic4Wideband Phone (VoLTE)6Randomized Phone Channel7Narrowband 2G + Packet Loss14

What did the model get wrong?

Ethio-ASR Multi (300M) Randomized Phone Channel   Example 6 of 20  ยท  speaker: Male

๐ŸŽง Listen โ€” Randomized Phone Channel
Phone-call audio with random extra distortions layered on top.
โœ… What was actually said
แˆ˜แ‰ธ แАแ‹ แ‹จแˆšแˆตแ‰ฐแŠซแŠจแˆˆแ‹ แˆตแˆซ แˆตแˆ‹แˆˆแ‰ฅแŠ แ‰ถแˆŽ แ‰ขแˆ†แŠ•แˆแŠ แŒฅแˆฉ แАแ‹
๐Ÿค– What this model heard
amh แˆแ‰ธ แАแ‹ แ‹จแˆšแˆตแ‰ฐแŠซแŠจแˆˆแ‹ แˆตแˆซแ‰ต แˆ‹แˆแˆˆแ‰ขแŠ• แ‰ฐแˆŽ แˆ†แА แˆˆ แŒฅแˆฉ แАแ‹
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
4
Correct
5
Wrong
0
Missed
2
Extra
44%
Words right
Out of 9 spoken words, this model got 4 right (44%). Errors: 5 words wrong; 2 extra.

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