Instructions to use davitlive5/Lapillus-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- NeMo
How to use davitlive5/Lapillus-v2 with NeMo:
import nemo.collections.asr as nemo_asr asr_model = nemo_asr.models.ASRModel.from_pretrained("davitlive5/Lapillus-v2") transcriptions = asr_model.transcribe(["file.wav"]) - Notebooks
- Google Colab
- Kaggle
Lapillus-v2
Lapillus-v2 is an Armenian speech recognition model created by David Ghazaryan, tuned for call-center and telephone audio while keeping Lapillus-v1's accuracy on clean speech.
It is a FastConformer hybrid RNN-T + CTC model (115M parameters), continued from
davitlive5/Lapillus-v1. It outputs text with punctuation and
capitalisation. The repository also contains an optional 6-gram Armenian language model for beam-search decoding.
Official ArmBench results
Independent evaluation by Metric AI lab on ArmBench-ASR (%, lower is better). "norm" = after text normalisation (case, punctuation, numbers); raw scores also count punctuation, capitalisation and number formatting.
| Dataset | WER | WER (norm) | CER | CER (norm) |
|---|---|---|---|---|
| Common Voice 26 | 7.02 | 4.34 | 1.42 | 0.99 |
| FLEURS | 12.83 | 5.43 | 2.97 | 1.82 |
| Infocom (May 2026) | 19.01 | 5.45 | 4.05 | 1.81 |
| Poems | 30.36 | 9.20 | 7.98 | 3.46 |
| Movies | 58.18 | 33.96 | 27.85 | 20.21 |
| Combined | 18.21 | 8.31 | 4.95 | 2.94 |
Most of the gap between raw and normalised scores comes from numbers (the model writes them as words, e.g. "ึีฝีกีถีฐีซีถีฃ" instead of "25") and punctuation. Film dialogue (music, effects, overlapping speech) remains the hardest domain.
Our evaluation
Normalised WER (%), lower is better. "Phone" sets are the official test audio passed through a fixed call-center simulation (background noise / babble / music from held-out files, room reverb, 300โ3400 Hz band, 8 kHz, G.711 ฮผ-law / A-law, packet loss).
| Test set | Lapillus-v1 (greedy) | Lapillus-v2 (beam 4 + LM) | Relative |
|---|---|---|---|
| ArmBench Common Voice 26 test (6892 utts) | 5.19 | 4.48 | โ14% |
| ArmBench FLEURS test (932 utts) | 6.70 | 5.73 | โ14% |
| Phone Common Voice 26 test (2000 utts) | 15.21 | 12.50 | โ18% |
| Phone FLEURS test (932 utts) | 12.42 | 9.86 | โ21% |
With the same beam + LM decoder, Lapillus-v1 scores 4.51 / 5.91 / 13.13 / 10.59, so the new weights are better on all four sets and the decoder adds most of the clean-speech gain.
Usage
# pip install "nemo_toolkit[asr]"
import nemo.collections.asr as nemo_asr
model = nemo_asr.models.ASRModel.from_pretrained("davitlive5/Lapillus-v2")
print(model.transcribe(["audio.wav"])[0].text) # 16 kHz mono (8 kHz phone audio: resample to 16 kHz)
Best accuracy: beam search + language model
from huggingface_hub import hf_hub_download
from omegaconf import open_dict
lm = hf_hub_download("davitlive5/Lapillus-v2", "lm/hy_tok6_lm.nemo")
cfg = model.cfg.decoding
with open_dict(cfg):
cfg.strategy = "malsd_batch"
cfg.beam.beam_size = 4
cfg.beam.return_best_hypothesis = True
cfg.beam.ngram_lm_model = lm
cfg.beam.ngram_lm_alpha = 0.3
model.change_decoding_strategy(cfg, decoder_type="rnnt")
print(model.transcribe(["audio.wav"])[0].text)
On a GPU this runs at about the same speed as greedy decoding (~0.13 s for a 7 s utterance on an RTX 5070). Split audio longer than about 30 seconds into shorter segments before transcribing.
Training
- Fine-tuning from Lapillus-v1 on about 87 h of public Eastern Armenian speech (OpenSLR SLR160, FLEURS train, Armenian audiobooks, VoxPopuli) with on-the-fly call-center augmentation: speed 0.9โ1.1, room impulse responses, MUSAN noise and music, 3โ6 speaker babble, telephone band-pass, 8 kHz resampling, G.711 codecs, packet loss and clipping. Loss 0.7 RNN-T + 0.3 CTC, AdamW lr 3e-5 with cosine schedule, EMA 0.999, 12k steps, bf16. Best EMA checkpoint at 10k steps.
- Weight averaging: the fine-tuned weights are averaged 50/50 with Lapillus-v1, which keeps v1's clean-speech accuracy and most of the telephone gain.
- Language model (
lm/hy_tok6_lm.nemo): token-level 6-gram KenLM on 8M Armenian sentences from FineWeb-2 plus training transcripts.
Every training sentence and LM sentence that matches an ArmBench test sentence, or shares any 7-word sequence with one, was removed. No Common Voice 26 test clips were used.
Limitations
- Eastern Armenian only. Expect more errors on Western Armenian.
- Tested on simulated telephone audio; real call recordings may differ.
- Accuracy drops on film dialogue, heavy overlapping speech, rare names and specialised terms. Numbers are written out as words.
License
CC-BY-4.0, derived from davitlive5/Lapillus-v1 and NVIDIA's
stt_hy_fastconformer_hybrid_large_pc (CC-BY-4.0).
Citation
@misc{ghazaryan2026lapillusv2,
title = {Lapillus-v2: Armenian Speech Recognition for Call Centers},
author = {Ghazaryan, David},
year = {2026},
url = {https://huggingface.co/davitlive5/Lapillus-v2}
}
- Downloads last month
- 31
Model tree for davitlive5/Lapillus-v2
Base model
nvidia/stt_hy_fastconformer_hybrid_large_pcEvaluation results
- WER (normalised, official ArmBench) on ArmBench-ASR Common Voice 26 testself-reported4.340
- CER (normalised, official ArmBench) on ArmBench-ASR Common Voice 26 testself-reported0.990
- WER (normalised, official ArmBench) on ArmBench-ASR FLEURS testself-reported5.430
- CER (normalised, official ArmBench) on ArmBench-ASR FLEURS testself-reported1.820
- WER (normalised, official ArmBench) on ArmBench-ASR combined (MCV26, FLEURS, Infocom, Poems, Movies)self-reported8.310
- CER (normalised, official ArmBench) on ArmBench-ASR combined (MCV26, FLEURS, Infocom, Poems, Movies)self-reported2.940