← All examples

ModelEthio-ASR Amharic (600M)11Ethio-ASR Multi (300M)13Ethio-ASR Multi (600M)9Ethio-ASR Multi (94M)8Meta MMS-1B19OmniASR-CTC (1B)0OmniASR-CTC (300M)0OmniASR-CTC (3B)0OmniASR-LLM (1B)0OmniASR-LLM (300M)0OmniASR-LLM (3B)0OpenAI Whisper-small20
Audio qualityClean Studio220Ambient Room Noise71Harsh Environment219Distant / Low-Volume Mic220Wideband Phone (VoLTE)18Randomized Phone Channel20Narrowband 2G + Packet Loss19

What did the model get wrong?

OpenAI Whisper-small Randomized Phone Channel   Example 1 of 20  ·  speaker: Female

🎧 Listen — Randomized Phone Channel
Phone-call audio with random extra distortions layered on top.
✅ What was actually said
ሁለት ዜሮ ሁለት አምስት አንድ ሁለት አንድ ሰባት ሁለት አራት ሰባት ስምንት ዜሮ ሰባት ዘጠኝ አራት
🤖 What this model heard
c c t hulet ዜሮ hulet thumb mist and hulet and savat hulet arad savats and mint ዜሮ savat jat and arad
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
2
Correct
17
Wrong
0
Missed
3
Extra
11%
Words right
Out of 19 spoken words, this model got 2 right (11%). Errors: 17 words wrong; 3 extra.

Try switching Audio quality — see the same model degrade as the audio gets noisier.