Parakeet CTC 0.6B English (en-US) - Speech-to-Text Collection

Overview

Automatic Speech Recognition (ASR) systems are widely used in applications such as voice command interfaces, automated subtitle generation for online media, and transcription of customer interactions for archival. Recent advances in deep learning have enabled speech-to-text models to transcribe audio to text in real time, achieving high accuracy even in adverse acoustic conditions, with robustness to noise and accents, and low word error rates (WER).

The Parakeet CTC 0.6B model represents a significant advancement in conversational AI, offering state-of-the-art English (en-US) speech recognition with 600 million parameters. Integrated within the NeMo framework, it provides developers with a highly accurate and efficient solution for building robust voice-enabled applications.

In this collection

This collection provides a comprehensive set of finetunable artifacts and a ready-to-use Parakeet NIM container, enabling flexible deployment and customization of the Parakeet CTC 0.6B model as a NVIDIA Inference Microservice (NIM). The collection includes:

1. Acoustic Model - Parakeet-CTC (FastConformer-CTC) base model for audio-to-text transcription
2. Language Model - N-gram language model for contextual word sequence prediction
3. Punctuation and Capitalization Model - Restores punctuation and proper capitalization to raw transcripts
4. Speaker Diarization Model (Sortformer) - Identifies and labels different speakers in audio (who spoke when)
5. Voice Activity Detection (VAD) - Detects speech segments in audio streams
6. Inverse Text Normalization (ITN) - Converts spoken form text to written form (e.g., "twenty twenty five" → "2025")

Each component can be used as provided or further finetuned to suit custom deployment needs, enabling domain-specific customization for various industries and use cases.

How does Parakeet CTC work?

The Parakeet CTC pipeline processes audio through multiple specialized components that work together to produce accurate, properly formatted transcriptions with speaker labels:

1. Voice Activity Detection (VAD)

Before processing begins, the Silero VAD model identifies segments of the audio that contain speech versus silence or background noise. This preprocessing step:

• Reduces computational overhead by filtering non-speech segments
• Improves noise robustness by focusing processing on relevant audio
• Enhances the ability to detect the beginning and end of speech segments for acurate endpointing.

2. Feature Extraction

The audio preprocessing stage converts raw analog signals into a machine-understandable format:

• Resampling: Converts audio to the appropriate sample rate
• Signal Preprocessing: Applies standardization and windowing techniques
• Spectrogram Transformation: Converts time-domain audio signals into frequency-domain representations that capture acoustic features

3. Acoustic Modeling with Parakeet-CTC

The core of the system is the Parakeet-CTC (FastConformer-CTC) acoustic model, which:

• Architecture: Uses an optimized Conformer architecture with 8x depthwise-separable convolutional downsampling
• Input Processing: Accepts spectrogram features and processes them through self-attention and convolution layers
• CTC Loss: Employs Connectionist Temporal Classification to align audio frames with text outputs
• Output: Produces log probability scores for vocabulary tokens at each time step

The FastConformer architecture combines the strengths of Transformers (global context via self-attention) with CNNs (local feature extraction), achieving superior accuracy with improved efficiency through linearly scalable attention mechanisms.

4. Language Modeling

The n-gram language model adds contextual understanding:

• Statistical Context: Calculates probability distributions over word sequences
• Error Correction: Helps correct acoustic model mistakes by considering likely word combinations
• Domain Adaptation: Can be customized with domain-specific vocabulary and patterns
• Architecture: 3-gram model trained with Kneser-Ney smoothing (4-gram for unified models)

5. Inverse Text Normalization (ITN)

ITN converts spoken-form text to written form using weighted finite-state transducers (WFST):

• Numbers: "twenty twenty five" → "2025"
• Measurements: "five feet" → "5 ft"
• Currency: "ten dollars" → "$10"
• Dates and times: "january first" → "January 1st"

The ITN model uses OpenFST finite state archives (.far) within the Sparrowhawk normalization engine.

6. Punctuation and Capitalization

The punctuation and capitalization model restores proper formatting:

• Adds periods, commas and question marks punctuation symbols
• Capitalizes proper nouns, sentence beginnings, etc
• Improves readability and downstream NLP task performance
• Uses context-aware deep learning models trained on properly formatted text

7. Speaker Diarization with Sortformer

The Sortformer diarization model identifies and labels different speakers in multi-speaker audio:

• Speaker Segmentation: Determines when different speakers are talking
• Speaker Labeling: Assigns consistent labels to each speaker throughout the audio
• Streaming Support: Provides real-time speaker diarization with low latency
• Multi-speaker Support: Handles up to 4 speakers simultaneously

Sortformer uses an innovative arrival-time ordering approach to resolve speaker permutation problems, making it particularly effective for conversational audio, meetings, and interviews. The model employs a FastConformer encoder with Transformer layers trained with Sort-Loss and Permutation Invariant Loss objectives.

Where to Get Started

Prerequisites

Runtime Environment:

• Riva 2.15.0 or higher
• Linux operating system
• NVIDIA GPU (H100, A100, or L40 recommended)

Development Tools:

• NeMo Framework for model finetuning
• NGC CLI for downloading artifacts
• Docker for containerized deployment

Deployment Workflow

1. Download the artifacts: Use NGC CLI to download all artifacts from this collection

   ngc registry model download-version <artifact>
   

2. Finetune for Your Domain (Optional): Customize models for your specific use case

   • Acoustic Model: Finetune on domain-specific audio (accents, terminology, acoustics) Tutorial
   • Language Model: Train on domain-specific text corpus for better contextual predictions Tutorial
   • Punctuation Model: Adapt to domain-specific formatting conventions
   • ITN Rules: Customize normalization rules for domain-specific entities Tutorial

3. Optimize and Deploy: Package customized models into a NIM

   • Use Riva build and deploy flow to convert and optimize models
   • Deploy as a microservice with Riva NIMs
   • Scale horizontally for production workloads

References

Academic Papers

1. Fast Conformer with Linearly Scalable Attention for Efficient Speech Recognition
2. Conformer: Convolution-augmented Transformer for Speech Recognition
3. NeMo Inverse Text Normalization: From Development To Production

Software & Tools

1. NeMo Framework - Training and finetuning toolkit
2. KenLM - Language model training
3. Google Sparrowhawk - Text normalization engine
4. Riva - Deployment and inference platform

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.

Terms of Use

By using this software or microservice, you are agreeing to the terms and conditions of the license and acceptable use policy.

Governing Terms

This trial is governed by the NVIDIA API Trial Terms of Service (found at https://assets.ngc.nvidia.com/products/api-catalog/legal/NVIDIA%20API%20Trial%20Terms%20of%20Service.pdf). The use of this model is governed by the AI Foundation Models Community License Agreement.

You are responsible for ensuring that your use of NVIDIA AI Foundation Models complies with all applicable laws.

Disclaimer

AI models generate responses and outputs based on complex algorithms and machine learning techniques, and those responses or outputs may be inaccurate or indecent. By testing this model, you assume the risk of any harm caused by any response or output of the model. Please do not upload any confidential information or personal data. Your use is logged for security.