Ported from upstream commit
675e786,
pinned 2026-05-10.
Validated against the NeMo reference at transcribe.cpp commit
42528dd
on 2026-05-10.
Offline English speech-to-text with punctuation and capitalization. A 1.1B-parameter FastConformer-XL encoder with a TDT/RNNT transducer decoder (the auxiliary CTC head from the upstream hybrid checkpoint is dropped at convert time). Not a streaming model and does not translate.
WER measured on the full LibriSpeech test-clean split (2620 utterances) with greedy TDT/RNN-T transducer decoding and no external LM. F32 reference baseline: 1.87%. NVIDIA's self-reported number on the same split is 1.82%.
If your audio isn't already 16 kHz mono WAV, convert it first:
ffmpeg -i input.mp3 -ar 16000 -ac 1 output.wav
See the transcribe.cpp model page for performance
numbers, numerical validation, and reproduction steps.
License
Inherited from the base model: CC-BY-4.0. See the
upstream model card for full terms.
Original Model Card
The section below is reproduced from
nvidia/parakeet-tdt_ctc-1.1b at commit
675e786 for offline reference. The upstream card is the
authoritative source.
parakeet-tdt_ctc-1.1b is an ASR model that transcribes speech with Punctuations and Capitalizations of English alphabet. This model is jointly developed by NVIDIA NeMo and Suno.ai teams.
It is an XXL version of Hybrid FastConformer [1] TDT-CTC [2] (around 1.1B parameters) model. This model has been trained with Local Attention and Global token hence this model can transcribe 11 hrs of audio in one single pass. And for reference this model can transcibe 90mins of audio in <16 sec on A100.
See the model architecture section and NeMo documentation for complete architecture details.
NVIDIA NeMo: Training
To train, fine-tune or play with the model you will need to install NVIDIA NeMo. We recommend you install it after you've installed latest PyTorch version.
pip install nemo_toolkit['all']
How to Use this Model
The model is available for use in the NeMo toolkit [3], and can be used as a pre-trained checkpoint for inference or for fine-tuning on another dataset.
Automatically instantiate the model
python
1import nemo.collections.asr as nemo_asr
2asr_model = nemo_asr.models.ASRModel.from_pretrained(model_name="nvidia/parakeet-tdt_ctc-1.1b")
This model accepts 16000 Hz mono-channel audio (wav files) as input.
Output
This model provides transcribed speech as a string for a given audio sample.
Model Architecture
This model uses a Hybrid FastConformer-TDT-CTC architecture. FastConformer [1] is an optimized version of the Conformer model with 8x depthwise-separable convolutional downsampling. You may find more information on the details of FastConformer here: Fast-Conformer Model.
Training
The NeMo toolkit [3] was used for finetuning this model for 20,000 steps over parakeet-tdt-1.1 model. This model is trained with this example script and this base config.
The tokenizers for these models were built using the text transcripts of the train set with this script.
Datasets
The model was trained on 36K hours of English speech collected and prepared by NVIDIA NeMo and Suno teams.
The training dataset consists of private subset with 27K hours of English speech plus 9k hours from the following public PnC datasets:
The performance of Automatic Speech Recognition models is measuring using Word Error Rate. Since this dataset is trained on multiple domains and a much larger corpus, it will generally perform better at transcribing audio in general.
The following tables summarizes the performance of the available models in this collection with the Transducer decoder. Performances of the ASR models are reported in terms of Word Error Rate (WER%) with greedy decoding.
Version
Tokenizer
Vocabulary Size
AMI
Earnings-22
Giga Speech
LS test-clean
LS test-other
SPGI Speech
TEDLIUM-v3
Vox Populi
Common Voice
1.23.0
SentencePiece Unigram
1024
15.94
11.86
10.19
1.82
3.67
2.24
3.87
6.19
8.69
These are greedy WER numbers without external LM. More details on evaluation can be found at HuggingFace ASR Leaderboard
Model Fairness Evaluation
As outlined in the paper "Towards Measuring Fairness in AI: the Casual Conversations Dataset", we assessed the parakeet-tdt_ctc-1.1b model for fairness. The model was evaluated on the CausalConversations-v1 dataset, and the results are reported as follows:
Gender Bias:
Gender
Male
Female
N/A
Other
Num utterances
19325
24532
926
33
% WER
12.81
10.49
13.88
23.12
Age Bias:
Age Group
(18-30)
(31-45)
(46-85)
(1-100)
Num utterances
15956
14585
13349
43890
% WER
11.50
11.63
11.38
11.51
(Error rates for fairness evaluation are determined by normalizing both the reference and predicted text, similar to the methods used in the evaluations found at https://github.com/huggingface/open_asr_leaderboard.)
NVIDIA Riva: Deployment
NVIDIA Riva, is an accelerated speech AI SDK deployable on-prem, in all clouds, multi-cloud, hybrid, on edge, and embedded.
Additionally, Riva provides:
World-class out-of-the-box accuracy for the most common languages with model checkpoints trained on proprietary data with hundreds of thousands of GPU-compute hours
Best in class accuracy with run-time word boosting (e.g., brand and product names) and customization of acoustic model, language model, and inverse text normalization
Streaming speech recognition, Kubernetes compatible scaling, and enterprise-grade support
License to use this model is covered by the CC-BY-4.0. By downloading the public and release version of the model, you accept the terms and conditions of the CC-BY-4.0 license.
