parakeet-tdt-0.6b-v3-gguf

GGUF conversions of [nvidia/parakeet-tdt-0.6b-v3](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3) for use with [transcribe.cpp](https://github.com/handy-computer/transcribe.cpp). Ported from upstream commit [6d590f7](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3/commit/6d590f7), pinned 2026-04-16. Validated against the NeMo reference at transcribe.cpp commit [bf0d0b7](https://github.com/handy-computer/transcribe.cpp/tree/bf0d0b7) on 2026-04-18. Offline multilingual speech-to-text covering 25 European languages. A 0.6B-parameter Conformer encoder with a TDT/RNNT transducer decoder. Takes a 16 kHz mono WAV and produces a transcript with optional token-level timestamps. Not a streaming model and does not translate. WER measured on the full LibriSpeech test-clean split (2620 utterances) with greedy transducer decoding and no external LM. F32 reference baseline: 1.95%. NVIDIA's self-reported number on the same split is 1.93%. 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 parakeet-tdt-0.6b-v3-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](https://github.com/handy-computer/transcribe.cpp/blob/main/docs/models/parakeet-tdt-0.6b-v3.md) for performance numbers, numerical validation, and reproduction steps. Inherited from the base model: **CC-BY-4.0**. See the [upstream model card](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3) for full terms. > The section below is reproduced from > [nvidia/parakeet-tdt-0.6b-v3](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3) at commit > `6d590f7` for offline reference. The upstream card is the > authoritative source. img { display: inline; } [](#model-architecture) `parakeet-tdt-0.6b-v3` is a 600-million-parameter multilingual automatic speech recognition (ASR) model designed for high-throughput speech-to-text transcription. It extends the [parakeet-tdt-0.6b-v2](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2) model by expanding language support from English to 25 European languages. The model automatically detects the language of the audio and transcribes it without requiring additional prompting. It is part of a series of models that leverage the [Granary](https://huggingface.co/datasets/nvidia/Granary) [1, 2] multilingual corpus as their primary training dataset. 🗣️ Try Demo here: https://huggingface.co/spaces/nvidia/parakeet-tdt-0.6b-v3 **Supported Languages:** Bulgarian (**bg**), Croatian (**hr**), Czech (**cs**), Danish (**da**), Dutch (**nl**), English (**en**), Estonian (**et**), Finnish (**fi**), French (**fr**), German (**de**), Greek (**el**), Hungarian (**hu**), Italian (**it**), Latvian (**lv**), Lithuanian (**lt**), Maltese (**mt**), Polish (**pl**), Portuguese (**pt**), Romanian (**ro**), Slovak (**sk**), Slovenian (**sl**), Spanish (**es**), Swedish (**sv**), Russian (**ru**), Ukrainian (**uk**) This model is ready for commercial/non-commercial use. `parakeet-tdt-0.6b-v3`'s key features are built on the foundation of its predecessor, [parakeet-tdt-0.6b-v2](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2), and include: * Automatic **punctuation** and **capitalization** * Accurate **word-level** and **segment-level** timestamps * **Long audio** transcription, supporting audio **up to 24 minutes** long with full attention (on A100 80GB) or up to 3 hours with local attention. * Released under a **permissive CC BY 4.0 license** For full details on the model architecture, training methodology, datasets, and evaluation results, check out the **[Technical Report](https://arxiv.org/abs/2509.14128)**. GOVERNING TERMS: Use of this model is governed by the [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/legalcode.en) license. 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](https://www.nvidia.com/en-us/ai-data-science/foundation-models/nemotron/)?<br> NVIDIA Developer [Nemotron](https://developer.nvidia.com/nemotron)<br> [NVIDIA Riva Speech](https://developer.nvidia.com/riva?sortBy=developer_learning_library%2Fsort%2Ffeatured_in.riva%3Adesc%2Ctitle%3Aasc#demos)<br> [NeMo Documentation](https://docs.nvidia.com/nemo-framework/user-guide/latest/nemotoolkit/asr/models.html)<br> *Figure 1: ASR WER comparison across different models. This does not include Punctuation and Capitalisation errors.* **Note 1:** The above evaluations are conducted for 24 supported languages, excluding Latvian since `seamless-m4t-v2-large` and `seamless-m4t-medium` do not support it. **Note 2:** Performance differences may be partly attributed to Portuguese variant differences - our training data uses European Portuguese while most benchmarks use Brazilian Portuguese. Global This model serves developers, researchers, academics, and industries building applications that require speech-to-text capabilities, including but not limited to: conversational AI, voice assistants, transcription services, subtitle generation, and voice analytics platforms. Huggingface [08/14/2025](https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3) **Architecture Type**: FastConformer-TDT **Network Architecture**: * This model was developed based on [FastConformer encoder](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/asr/models.html#fast-conformer) architecture[3] and TDT decoder[4] * This model has 600 million model parameters. **Input Type(s):** 16kHz Audio **Input Format(s):** `.wav` and `.flac` audio formats **Input Parameters:** 1D (audio signal) **Other Properties Related to Input:** Monochannel audio **Output Type(s):** Text **Output Format:** String **Output Parameters:** 1D (text) **Other Properties Related to Output:** Punctuations and Capitalizations included. Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. 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. For more information, refer to the [NeMo documentation](https://docs.nvidia.com/deeplearning/nemo/user-guide/docs/en/main/asr/models.html#fast-conformer). To train, fine-tune or play with the model you will need to install [NVIDIA NeMo](https://github.com/NVIDIA/NeMo). We recommend you install it after you've installed latest PyTorch version. pip install -U nemo_toolkit['asr'] The model is available for use in the NeMo toolkit [5], and can be used as a pre-trained checkpoint for inference or for fine-tuning on another dataset. import nemo.collections.asr as nemo_asr asr_model = nemo_asr.models.ASRModel.from_pretrained(model_name="nvidia/parakeet-tdt-0.6b-v3") First, let's get a sample wget https://dldata-public.s3.us-east-2.amazonaws.com/2086-149220-0033.wav Then simply do: output = asr_model.transcribe(['2086-149220-0033.wav']) print(output[0].text) To transcribe with timestamps: output = asr_model.transcribe(['2086-149220-0033.wav'], timestamps=True) word_timestamps = output[0].timestamp['word'] # word level timestamps for first sample segment_timestamps = output[0].timestamp['segment'] # segment level timestamps char_timestamps = output[0].timestamp['char'] # char level timestamps for stamp in segment_timestamps: print(f"{stamp['start']}s - {stamp['end']}s : {stamp['segment']}") asr_model.change_attention_model(self_attention_model="rel_pos_local_attn", att_context_size=[256, 256]) output = asr_model.transcribe(['2086-149220-0033.wav']) print(output[0].text) To use parakeet models in streaming mode use this [script](https://github.com/NVIDIA/NeMo/blob/main/examples/asr/asr_chunked_inference/rnnt/speech_to_text_streaming_infer_rnnt.py) as shown below: python NeMo/main/examples/asr/asr_chunked_inference/rnnt/speech_to_text_streaming_infer_rnnt.py \ pretrained_name="nvidia/parakeet-tdt-0.6b-v3" \ model_path=null \ audio_dir="<optional path to folder of audio files>" \ dataset_manifest="<optional path to manifest>" \ output_filename="<optional output filename>" \ right_context_secs=2.0 \ chunk_secs=2 \ left_context_secs=10.0 \ batch_size=32 \ clean_groundtruth_text=False NVIDIA NIM for v2 parakeet model is available at [https://build.nvidia.com/nvidia/parakeet-tdt-0_6b-v2](https://build.nvidia.com/nvidia/parakeet-tdt-0_6b-v2). **Runtime Engine(s):** * NeMo 2.4 **Supported Hardware Microarchitecture Compatibility:** * NVIDIA Ampere * NVIDIA Blackwell * NVIDIA Hopper * NVIDIA Volta **[Preferred/Supported] Operating System(s):** - Linux **Hardware Specific Requirements:** Atleast 2GB RAM for model to load. The bigger the RAM, the larger audio input it supports. Current version: `parakeet-tdt-0.6b-v3`. Previous versions can be [accessed](https://huggingface.co/collections/nvidia/parakeet-659711f49d1469e51546e021) here. This model was trained using the NeMo toolkit [5], following the strategies below: - Initialized from a CTC multilingual checkpoint pretrained on the Granary dataset \[1] \[2]. - Trained for 150,000 steps on 128 A100 GPUs. - Dataset corpora and languages were balanced using a temperature sampling value of 0.5. - Stage 2 fine-tuning was performed for 5,000 steps on 4 A100 GPUs using approximately 7,500 hours of high-quality, human-transcribed data of NeMo ASR Set 3.0. Training was conducted using this [example script](https://github.com/NVIDIA/NeMo/blob/main/examples/asr/asr_transducer/speech_to_text_rnnt_bpe.py) and [TDT configuration](https://github.com/NVIDIA/NeMo/blob/main/examples/asr/conf/fastconformer/hybrid_transducer_ctc/fastconformer_hybrid_tdt_ctc_bpe.yaml). During the training, a unified SentencePiece Tokenizer \[6] with a vocabulary of **8,192 tokens** was used. The unified tokenizer was constructed from the training set transcripts using this [script](https://github.com/NVIDIA/NeMo/blob/main/scripts/tokenizers/process_asr_text_tokenizer.py) and was optimized across all 25 supported languages. The model was trained on the combination of [Granary dataset's ASR subset](https://huggingface.co/datasets/nvidia/Granary) and in-house dataset NeMo ASR Set 3.0: - 10,000 hours from human-transcribed NeMo ASR Set 3.0, including: - LibriSpeech (960 hours) - Fisher Corpus - National Speech Corpus Part 1 - VCTK - Europarl-ASR - Multilingual LibriSpeech - Mozilla Common Voice (v7.0) - AMI - 660,000 hours of pseudo-labeled data from Granary \[1] \[2], including: - [YTC](https://huggingface.co/datasets/FBK-MT/mosel) \[7] - [MOSEL](https://huggingface.co/datasets/FBK-MT/mosel) \[8] - [YODAS](https://huggingface.co/datasets/espnet/yodas-granary) \[9] All transcriptions preserve punctuation and capitalization. The Granary dataset will be made publicly available after presentation at Interspeech 2025. **Data Collection Method by dataset** * Hybrid: Automated, Human **Labeling Method by dataset** * Hybrid: Synthetic, Human **Properties:** * Noise robust data from various sources * Single channel, 16kHz sampled data For multilingual ASR performance evaluation: - Fleurs [10] - MLS [11] - CoVoST [12] For English ASR performance evaluation: - Hugging Face Open ASR Leaderboard [13] datasets **Data Collection Method by dataset** * Human **Labeling Method by dataset** * Human **Properties:** * All are commonly used for benchmarking English ASR systems. * Audio data is typically processed into a 16kHz mono channel format for ASR evaluation, consistent with benchmarks like the [Open ASR Leaderboard](https://huggingface.co/spaces/hf-audio/open_asr_leaderboard). The tables below summarizes the WER (%) using a Transducer decoder with greedy decoding (without an external language model): **Note:** WERs are calculated after removing Punctuation and Capitalization from reference and predicted text. Additional evaluation details are available on the [Hugging Face ASR Leaderboard](https://huggingface.co/spaces/hf-audio/open_asr_leaderboard).[13] Performance across different Signal-to-Noise Ratios (SNR) using MUSAN music and noise samples [14]: [1] [Granary: Speech Recognition and Translation Dataset in 25 European Languages](https://arxiv.org/abs/2505.13404) [2] [NVIDIA Granary Dataset Card](https://huggingface.co/datasets/nvidia/Granary) [3] [Fast Conformer with Linearly Scalable Attention for Efficient Speech Recognition](https://arxiv.org/abs/2305.05084) [4] [Efficient Sequence Transduction by Jointly Predicting Tokens and Durations](https://arxiv.org/abs/2304.06795) [5] [NVIDIA NeMo Toolkit](https://github.com/NVIDIA/NeMo) [6] [Google Sentencepiece Tokenizer](https://github.com/google/sentencepiece) [7] [Youtube-Commons](https://huggingface.co/datasets/PleIAs/YouTube-Commons) [8] [MOSEL: 950,000 Hours of Speech Data for Open-Source Speech Foundation Model Training on EU Languages](https://arxiv.org/abs/2410.01036) [9] [YODAS: Youtube-Oriented Dataset for Audio and Speech](https://arxiv.org/pdf/2406.00899) [10] [FLEURS: Few-shot Learning Evaluation of Universal Representations of Speech](https://arxiv.org/abs/2205.12446) [11] [MLS: A Large-Scale Multilingual Dataset for Speech Research](https://arxiv.org/abs/2012.03411) [12] [CoVoST 2 and Massively Multilingual Speech-to-Text Translation](https://arxiv.org/abs/2007.10310) [13] [HuggingFace ASR Leaderboard](https://huggingface.co/spaces/hf-audio/open_asr_leaderboard) [14] [MUSAN: A Music, Speech, and Noise Corpus](https://arxiv.org/abs/1510.08484) **Engine**: * NVIDIA NeMo **Test Hardware**: * NVIDIA A10 * NVIDIA A100 * NVIDIA A30 * NVIDIA H100 * NVIDIA L4 * NVIDIA L40 * NVIDIA Turing T4 * NVIDIA Volta V100 NVIDIA believes Trustworthy AI is a shared responsibility and we have estab