Property	Value
Base model	nvidia/parakeet-tdt-0.6b-v3
Architecture	FastConformer-TDT (600M params)
Language	Polish (pl)
Input	16 kHz mono audio
Output	Polish text with punctuation and capitalization
License	CC-BY-4.0

Evaluation Results

Evaluated on Common Voice 17.0 Polish (raw text, no normalization):

Split	WER	CER	Samples
Validation	6.07%	--	--
Test	11.81%	2.72%	9,230

Training

Fine-tuned on a curated subset of the BIGOS v2 benchmark, filtered to retain only sources with proper casing and punctuation:

Common Voice 15 -- 19,119 human-recorded Polish speech samples
M-AILABS / LibriVox -- 11,834 read Polish audiobook samples
PolyAI Minds14 -- 462 Polish banking dialog samples
Total training set: ~31,415 samples

Validation uses the BIGOS v2 validation split (same source filtering). Test evaluation uses Common Voice 17.0 Polish (independent test set).

Training Configuration

Parameter	Value
Optimizer	AdamW
Learning rate	5e-5 (cosine annealing)
Warmup	10% of total steps
Batch size	32
Precision	bf16-mixed
Gradient clipping	1.0
Early stopping	10 epochs patience on val WER
Best epoch	21

Usage

Installation

pip install nemo_toolkit[asr]

Transcribe Audio

python

import nemo.collections.asr as nemo_asr

# Load model
asr_model = nemo_asr.models.ASRModel.from_pretrained(
    model_name="yuriyvnv/parakeet-tdt-0.6b-polish"
)

# Transcribe
output = asr_model.transcribe(["audio.wav"])
print(output[0].text)

Transcribe with Timestamps

python

output = asr_model.transcribe(["audio.wav"], timestamps=True)

for stamp in output[0].timestamp["segment"]:
    print(f"{stamp['start']:.1f}s - {stamp['end']:.1f}s : {stamp['segment']}")

Long-Form Audio

For audio longer than 24 minutes, enable local attention:

python

asr_model.change_attention_model(
    self_attention_model="rel_pos_local_attn",
    att_context_size=[256, 256],
)
output = asr_model.transcribe(["long_audio.wav"])

Intended Use

This model is designed for transcribing Polish speech to text. It works best on:

Read speech and conversational Polish
Audio recorded at 16 kHz or higher
Segments up to 24 minutes (or longer with local attention enabled)

Limitations

Training data is sourced from read speech (audiobooks, Common Voice read prompts) and short banking dialogs; performance may differ on spontaneous or heavily accented speech
The model preserves punctuation and capitalization as seen in training data
Not suitable for real-time streaming without additional configuration