---

license: other

license_name: nvidia-open-model-license

license_link: https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/

base_model: nvidia/nemotron-speech-streaming-en-0.6b

base_model_relation: quantized

library_name: transcribe.cpp

pipeline_tag: automatic-speech-recognition

language:

- en

tags:

- gguf

- transcribe.cpp

- asr

- speech-to-text

- parakeet

- conformer

- rnnt

- streaming

- cache-aware

transcribe_cpp:

wer_librispeech_test_clean:

f32: 2.31

f16: 2.31

q8_0: 2.31

q6_k: 2.29

q5_k_m: 2.34

q4_k_m: 2.38

rtf_m4_max:

metal: 154.5

cpu: 32

rtf_ryzen_4750u:

vulkan: 13

cpu: 7.5

streaming: true

translate: false

lang_detect: false

timestamps: token

---

nemotron-speech-streaming-en-0.6b: transcribe.cpp GGUF

GGUF conversions of nvidia/nemotron-speech-streaming-en-0.6b for use

with transcribe.cpp.

Ported from upstream commit

ef3bf40,

pinned 2026-05-11.

Validated against the NeMo reference at transcribe.cpp commit

12f1076

on 2026-05-11.

Offline English speech-to-text with punctuation and capitalization. A 0.6B-parameter cache-aware streaming FastConformer encoder with an RNN-T transducer decoder. Runs in offline mode. The encoder preserves the upstream att_context_size=[70, 13] (1.12s) cache-aware attention mask end-to-end. transcribe.cpp does not yet expose a streaming session API.

Downloads

| Quantization | Download | Size | WER (LibriSpeech test-clean, offline) |

| --- | --- | ---: | ---: |

| F32 | nemotron-speech-streaming-en-0.6b-F32.gguf | 2.30 GB | 2.31% |

| F16 | nemotron-speech-streaming-en-0.6b-F16.gguf | 1.16 GB | 2.31% |

| Q8_0 | nemotron-speech-streaming-en-0.6b-Q8_0.gguf | 696 MB | 2.31% |

| Q6_K | nemotron-speech-streaming-en-0.6b-Q6_K.gguf | 573 MB | 2.29% |

| Q5_K_M | nemotron-speech-streaming-en-0.6b-Q5_K_M.gguf | 514 MB | 2.34% |

| Q4_K_M | nemotron-speech-streaming-en-0.6b-Q4_K_M.gguf | 453 MB | 2.38% |

WER measured on the full LibriSpeech test-clean split (2620 utterances) with greedy RNN-T decoding. F32 reference baseline: 2.31%. NVIDIA's self-reported number on the same split at att_context_size=[70, 13] (1.12s chunk, w/o PnC) is 2.32%.

Usage

Build transcribe.cpp from source:

git clone git@github.com:handy-computer/transcribe.cpp.git
cd transcribe.cpp
cmake -B build && cmake --build build

Run on a 16 kHz mono WAV:

build/bin/transcribe-cli \
  -m nemotron-speech-streaming-en-0.6b-Q8_0.gguf \
  input.wav

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: NVIDIA Open Model License. See the

upstream model card for full terms.

---

Original Model Card

> The section below is reproduced from

> nvidia/nemotron-speech-streaming-en-0.6b at commit

> ef3bf40 for offline reference. The upstream card is the

> authoritative source.

Nemotron ASR Streaming

<style>

h1, h2, h3, h4, h5, h6 {

color: #76b900; / NVIDIA green /

font-weight: 700;

}

hr {

border: none;

border-top: 1px solid #e5e7eb;

margin: 2rem 0;

}

/ Improve list spacing /

ul, ol {

margin-top: 0.5rem;

margin-bottom: 0.5rem;

}

/ Badge alignment consistency /

img {

display: inline;

vertical-align: middle;

}

</style>


| 

| 

> [!Note]

> March 12, 2026: nemotron-asr-streaming was released with updated checkpoint (trained on larger corpora).

>

> For the older checkpoint released in January 2026, please refer to the nemotron-speech-streaming-jan2026 branch.

Nemotron-ASR-Streaming is an English, streaming Automatic Speech Recognition (ASR) engineered to deliver high-quality English transcription across both low-latency streaming and high-throughput batch workloads. Developed by NVIDIA, this 600M parameter model transcribes speech into text with native support for punctuation and capitalization and offers runtime flexibility with configurable chunk sizes, including 80ms, 160ms, 560ms, and 1120ms.

By leveraging the state-of-the-art Cache-Aware FastConformer-RNNT architecture, the model eliminates redundant overlapping computations common in traditional "buffered" streaming. This allows it to process only new audio chunks while reusing cached encoder context, significantly improving computational efficiency and minimizing end-to-end delay without sacrificing accuracy.

It transcribes speech into the English alphabet, spaces, and apostrophes, with full support for punctuation and capitalization. Trained on the ASRSet, a massive dataset of approximately 250,000 hours of US English (en-US) speech, it is engineered to perform across diverse and challenging acoustic conditions.

Why Choose nvidia/nemotron-asr-streaming?

• Native Streaming Architecture: Cache-aware design enables efficient processing of continuous audio streams, designed and optimized for low-latency voice agent applications interaction.
• Improved Operational Efficiency: Delivers superior throughput compared to traditional buffered streaming approaches. This allows for a higher number of parallel streams within the same GPU memory constraints, directly reducing operational costs for production environments.
• Dynamic Runtime Flexibility: Enables you to choose the optimal operating point on the latency-accuracy Pareto curve at inference time. No re-training is required to adjust for different use-case requirements.
• Punctuation & Capitalization: Built-in support for punctuation and capitalization in output text

<figure align="center">

<img src="figures/results_wer_and_scaling.png" width="1250" />

<figcaption>

Nemotron-asr-streaming, outperforming Jan26 version, allows users to choose the optimal operating point on the latency-accuracy pareto curve at inference time, without requiring any re-training. Further, the cache-aware streaming mechanism scales much better than buffered streaming approaches, consistently outperforming production models like <a href="https://build.nvidia.com/nvidia/parakeet-ctc-1_1b-asr"> parakeet-ctc-1_1b-asr </a> across chunk sizes.

</figcaption>

</figure>

This model consists of a cache-aware streaming 🦜 Parakeet (FastConformer) encoder with an RNN-T decoder. It is designed for real-time speech-to-text applications where low latency is critical, such as voice assistants, live captioning, and conversational AI systems. Unlike traditional "buffered" streaming, the cache-aware architecture enables continuous transcription by processing only new audio chunks while reusing cached encoder context. This significantly improves computational efficiency and minimizes end-to-end delay without sacrificing accuracy.

🗣️ Experience Nemotron-Speech-Streaming-En-0.6b in action here: https://huggingface.co/spaces/nvidia/nemotron-speech-streaming-en-0.6b

This model is ready for commercial/non-commercial use.

Read more about the model in the dev blog and check out the paper.

<div align="center">

<img src="figures/inference_pipeline.png" width="750" />

</div>

License/Terms of Use:

Governing Terms: Use of the model is governed by the NVIDIA Open Model License Agreement.

Deployment Geography:

Global <br>

Use Case: <br>

This model is for transcription of English audio. <br>

Release Date: <br>

• Build.Nvidia.com [03/13/2026] via https://build.nvidia.com/nvidia/nemotron-asr-streaming/ <br>
• Hugging Face [03/13/2026] via https://huggingface.co/nvidia/nemotron-speech-streaming-en-0.6b <br>

(older checkpoint [01/05/2026] via https://huggingface.co/nvidia/nemotron-speech-streaming-en-0.6b/tree/nemotron-speech-streaming-jan2026) <br>

• NGC [03/13/2026] via https://catalog.ngc.nvidia.com/orgs/nvidia/collections/nemotron-asr-streaming <br>

Explore more from NVIDIA:

For documentation, deployment guides, enterprise-ready APIs, and the latest open models—including Nemotron and other cutting-edge speech, translation, and generative AI—visit the NVIDIA Developer Portal at developer.nvidia.com.

Join the community to access tools, support, and resources to accelerate your development with NVIDIA's NeMo, Riva, NIM, and foundation models.<br>

What is Nemotron?<br>

NVIDIA Developer Nemotron<br>

NVIDIA Riva Speech<br>

NeMo Documentation<br>

Also, check out the following NVIDIA speech models that extend the capabilities of Nemotron-Speech-Streaming:

• Multitalker Parakeet Streaming - https://huggingface.co/nvidia/multitalker-parakeet-streaming-0.6b-v1
• Parakeet Realtime EOU - https://huggingface.co/nvidia/parakeet_realtime_eou_120m-v1

Access Model Inference and Examples:

• NVIDIA NIM
• Deploy the Nemotron Speech ASR endpoint on Modal
• Build local voice agent using Daily’s framework

Model Architecture

Architecture Type: FastConformer-CacheAware-RNNT

The model is based on the Cache-Aware [1] FastConformer [2] architecture with 24 encoder layers and an RNNT (Recurrent Neural Network Transducer) decoder. The cache-aware streaming design enables efficient processing of audio in chunks while maintaining context from previous frames. Unlike buffered inference, this model maintains caches for all encoder self-attention and convolution layers. This enables reuse of hidden states at every streaming step, where cached activations eliminate redundant computations. As a result, there are no overlapping computations; each processed frame is strictly non-overlapping.

The caching schema of self-attention and convolution layers for consecutive chunks is as follows. For more details, please refer to [1].

<div align="center">

<img src="figures/caching_schema.png" width="750" />

</div>

Network Architecture:

• Encoder: Cache-Aware FastConformer with 24 layers
• Decoder: RNNT (Recurrent Neural Network Transducer)
• Parameters: 600M

NVIDIA NeMo

To train, fine-tune or perform inference with this model, you will need to install NVIDIA NeMo[4]. We recommend you install it after you've installed Cython and latest PyTorch version.

apt-get update && apt-get install -y libsndfile1 ffmpeg
pip install Cython packaging
pip install git+https://github.com/NVIDIA/NeMo.git@main#egg=nemo_toolkit[asr]

How to Use this Model

The model is available for use in the NeMo Framework, and can be used as a pre-trained checkpoint for inference or for fine-tuning on another dataset.

Loading the Model

import nemo.collections.asr as nemo_asr
asr_model = nemo_asr.models.ASRModel.from_pretrained(model_name="nvidia/nemotron-speech-streaming-en-0.6b")

Streaming Inference

You can use the cache-aware streaming inference script from NeMo - NeMo/examples/asr/asr_cache_aware_streaming/speech_to_text_cache_aware_streaming_infer.py

cd NeMo
python examples/asr/asr_cache_aware_streaming/speech_to_text_cache_aware_streaming_infer.py \
    model_path=<model_path> \
    dataset_manifest=<dataset_manifest> \ 
    batch_size=<batch_size> \
    att_context_size="[70,13]" \ #set the second value to the desired right context from {0,1,6,13}
    output_path=<output_folder> 

You can also run streaming inference through the pipeline method, which uses NeMo/examples/asr/conf/asr_streaming_inference/cache_aware_rnnt.yaml configuration file to build end‑to‑end workflows with punctuation and capitalization (PnC), inverse text normalization (ITN), and translation support.

from nemo.collections.asr.inference.factory.pipeline_builder import PipelineBuilder
from omegaconf import OmegaConf

# Path to the cache aware config file downloaded from above link
cfg_path = 'cache_aware_rnnt.yaml'
cfg = OmegaConf.load(cfg_path)

# Pass the paths of all the audio files for inferencing
audios = ['/path/to/your/audio.wav']

# Create the pipeline object and run inference
pipeline = PipelineBuilder.build_pipeline(cfg)
output = pipeline.run(audios)

# Print the output
for entry in output:
  print(entry['text'])

_______

Setting up Streaming Configuration

Latency is defined by the att_context_size param, where att_context_size = {num_frames_left_context, num_frame_right_context}, all measured in 80ms frames:

• [70, 0]: Chunk size = 1 (1 × 80ms = 0.08s)
• [70, 1]: Chunk size = 2 (2 × 80ms = 0.16s)
• [70, 6]: Chunk size = 7 (7 × 80ms = 0.56s)
• [70, 13]: Chunk size = 14 (14 × 80ms = 1.12s)

Here, chunk size = current frame + right context; each chunk is processed in non-overlapping fashion.

Input

* Input Type(s): Audio <br>

* Input Format(s): wav <br>

* Input Parameters: One-Dimensional (1D) <br>

* Other Properties Related to Input: Maximum Length in seconds specific to GPU Memory, No Pre-Processing Needed, Mono channel is required. By leveraging NVIDIA’s hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions. <br>

Output

* Output Type(s): Text String in English <br>

* Output Format(s): String <br>

* Output Parameters: One-Dimensional (1D) <br>

* Other Properties Related to Output: No Maximum Character Length, transcribe punctuation and capitalization. By leveraging NVIDIA’s hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions. <br>

Try via API — No Setup Required

Transcribe audio using the hosted NVIDIA NIM API on build.nvidia.com — no local GPU, Docker, or model download required.

1. Get a free API key: Open Nemotron ASR Streaming and choose Get API Key.

2. Install the Riva client:

pip install nvidia-riva-client

3. Transcribe an audio file (offline / whole-file):

import riva.client

auth = riva.client.Auth(
    uri="grpc.nvcf.nvidia.com:443",
    use_ssl=True,
    metadata_args=[
        ["function-id", "bb0837de-8c7b-481f-9ec8-ef5663e9c1fa"],
        ["authorization", "Bearer nvapi-YOUR_API_KEY"],
    ],
)

asr_service = riva.client.ASRService(auth)

with open("audio.wav", "rb") as f:
    audio = f.read()

config = riva.client.RecognitionConfig(
    language_code="en-US",
    max_alternatives=1,
    enable_automatic_punctuation=True,
    enable_word_time_offsets=True,
)

response = asr_service.offline_recognize(audio, config)
print(response.results[0].alternatives[0].transcript)

Or use the CLI:

git clone https://github.com/nvidia-riva/python-clients.git
export NVIDIA_API_KEY="nvapi-YOUR_API_KEY"

python python-clients/scripts/asr/transcribe_file_offline.py \
    --server grpc.nvcf.nvidia.com:443 --use-ssl \
    --metadata function-id "bb0837de-8c7b-481f-9ec8-ef5663e9c1fa" \
    --metadata "authorization" "Bearer $NVIDIA_API_KEY" \
    --language-code en-US \
    --word-time-offsets --automatic-punctuation \
    --input-file audio.wav

> Note: For streaming, low-latency, and advanced options, use the API Reference tab on the Nemotron ASR Streaming model page. The hosted API typically accepts 16-bit mono audio in WAV, OGG, or OPUS; confirm details in the API reference.

Datasets

Training Datasets

The majority of the training data comes from NVIDIA Riva ASR training set (250k hours) and the English portion of the Granary dataset [3]:

• YouTube-Commons (YTC) (109.5k hours)
• YODAS2 (102k hours)
• Mosel (14k hours)
• LibriLight (49.5k hours)

In addition, the following datasets were used:

• Librispeech 960 hours
• Fisher Corpus
• Switchboard-1 Dataset
• WSJ-0 and WSJ-1
• National Speech Corpus (Part 1, Part 6)
• VCTK
• VoxPopuli (EN)
• Europarl-ASR (EN)
• Multilingual Librispeech (MLS EN)
• Mozilla Common Voice (v11.0)
• Mozilla Common Voice (v7.0)
• Mozilla Common Voice (v4.0)
• People Speech
• AMI

Data Modality: Audio and text

Audio Training Data Size: 530k hours

Data Collection Method: Human - All audios are human recorded

Labeling Method: Hybrid (Human, Synthetic) - Some transcripts are generated by ASR models, while some are manually labeled

Evaluation Datasets

The model was evaluated on the HuggingFace ASR Leaderboard datasets:

• AMI
• Earnings22
• Gigaspeech
• LibriSpeech test-clean
• LibriSpeech test-other
• SPGI Speech
• TEDLIUM
• VoxPopuli

Performance

ASR Performance (w/o PnC)

ASR performance is measured using the Word Error Rate (WER). Both ground-truth and predicted texts are processed using whisper-normalizer version 0.1.12.

The following tables show the WER on the HuggingFace OpenASR leaderboard datasets:

Word Error Rate (WER) for chunk size of 1.12s

| | Average | AMI | Earnings22 | Gigaspeech | LS-test-clean | LS-test-other | SPGI | TEDLIUM | VoxPopuli |

| :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |

| WER (%) | 6.93 | 11.73 | 12.52 | 9.66 | 2.32 | 4.84 | 2.97 | 3.50 | 7.91 |

WER for chunk size of 0.56s

| | Average | AMI | Earnings22 | Gigaspeech | LS-test-clean | LS-test-other | SPGI | TEDLIUM | VoxPopuli |

| :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |

| WER (%) | 7.07 | 11.88 | 12.82 | 9.78 | 2.46 | 5.07 | 3.03 | 3.54 | 8.00 |

WER for chunk size of 0.16s

| | Average | AMI | Earnings22 | Gigaspeech | LS-test-clean | LS-test-other | SPGI | TEDLIUM | VoxPopuli |

| :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |

| WER (%) | 7.67 | 14.71 | 13.01 | 10.34 | 2.56 | 5.57 | 3.25 | 3.77 | 8.18 |

WER for chunk size of 0.08s

| | Average | AMI | Earnings22 | Gigaspeech | LS-test-clean | LS-test-other | SPGI | TEDLIUM | VoxPopuli |

| :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |

| WER (%) | 8.43 | 18.29 | 13.16 | 11.17 | 2.80 | 6.01 | 3.43 | 4.10 | 8.46 |

Software Integration

Runtime Engine: NeMo 25.11, Riva 2.25.0 or higher

Supported Hardware Microarchitecture Compatibility:

• NVIDIA Ampere
• NVIDIA Blackwell
• NVIDIA Hopper
• NVIDIA Volta

Test Hardware:

• NVIDIA V100
• NVIDIA A100
• NVIDIA A6000
• DGX Spark

Preferred/Supported Operating System(s): Linux

Ethical Considerations

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.

References

[1] Stateful Conformer with Cache-based Inference for Streaming Automatic Speech Recognition

[2] Fast Conformer with Linearly Scalable Attention for Efficient Speech Recognition

[3] NVIDIA Granary

[4] NVIDIA NeMo Framework