A monolingual automatic speech recognition model for
Amharic
, fine-tuned from
openai/whisper-large-v3
. Part of
BuzzASR
,
a suite of 102 language-specialized ASR models (Findings of EMNLP 2026).
This model uses
full fine-tuning (native per-language tokenizer replacement + text multitask fine-tuning)
.
Results (normalized CER / WER, %)
Test set
CER
WER
Whisper-large-v3 (zero-shot) CER
FLEURS
11.8
33.87
193.45
Common Voice 25
0.02
0.08
187.69
Combined
9.07
26.13
191.35
~21.1x CER reduction over Whisper zero-shot on the combined test set.
amharic huggingface.co is an AI model on huggingface.co that provides amharic's model effect (), which can be used instantly with this BuzzASR amharic model. huggingface.co supports a free trial of the amharic model, and also provides paid use of the amharic. Support call amharic model through api, including Node.js, Python, http.
amharic huggingface.co is an online trial and call api platform, which integrates amharic's modeling effects, including api services, and provides a free online trial of amharic, you can try amharic online for free by clicking the link below.
BuzzASR amharic online free url in huggingface.co:
amharic is an open source model from GitHub that offers a free installation service, and any user can find amharic on GitHub to install. At the same time, huggingface.co provides the effect of amharic install, users can directly use amharic installed effect in huggingface.co for debugging and trial. It also supports api for free installation.