← All examples

ModelEthio-ASR Amharic (600M)3Ethio-ASR Multi (300M)10Ethio-ASR Multi (600M)3Ethio-ASR Multi (94M)6Meta MMS-1B9OmniASR-CTC (1B)2OmniASR-CTC (300M)7OmniASR-CTC (3B)3OmniASR-LLM (1B)5OmniASR-LLM (300M)7OmniASR-LLM (3B)2OpenAI Whisper-small9
Audio qualityClean Studio7Ambient Room Noise7Harsh Environment7Distant / Low-Volume Mic7Wideband Phone (VoLTE)7Randomized Phone Channel7Narrowband 2G + Packet Loss9

What did the model get wrong?

OmniASR-CTC (300M) Harsh Environment   Example 6 of 20  Β·  speaker: Male

🎧 Listen β€” Harsh Environment
Very heavy background noise β€” much harder than typical audio.
βœ… What was actually said
αˆ˜α‰Έ αŠα‹ α‹¨αˆšαˆ΅α‰°αŠ«αŠ¨αˆˆα‹ ሡራ αˆ΅αˆ‹αˆˆα‰₯ኝ α‰ΆαˆŽ α‰’αˆ†αŠ•αˆαŠ αŒ₯ሩ αŠα‹
πŸ€– What this model heard
αˆ˜α‰ΈαŠ• α‹¨αˆšαˆ΅α‰°αŠ«αŠ¨αˆˆα‹ αˆ΅αˆ«αˆ΅αˆ‹αˆˆα‰’αŠ• α‰°α‹ˆαˆŽα‰₯αˆ†αŠ•αˆŠ αŠ•αŒ₯ሩ αŠα‹
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
2
Correct
4
Wrong
3
Missed
0
Extra
22%
Words right
Out of 9 spoken words, this model got 2 right (22%). Errors: 4 words wrong; 3 missed.

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