> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/nemo/curator/_mcp/server. # Process Data ## Docs - [Overview](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/asr-inference.md): Perform automatic speech recognition using NeMo Framework models with GPU acceleration and batch processing - [NeMo ASR Models](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/asr-inference/nemo-models.md): Guide to using NeMo Framework's pretrained ASR models for speech recognition in audio curation pipelines - [Overview](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/quality-assessment.md): Assess and filter audio quality using transcription accuracy metrics, duration analysis, and custom quality measures - [WER Filtering](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/quality-assessment/wer-filtering.md): Filter audio samples based on Word Error Rate thresholds to ensure high-quality transcription accuracy - [Duration Filtering](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/quality-assessment/duration-filtering.md): Filter audio samples by duration ranges, speech rate metrics, and temporal characteristics for optimal dataset quality - [Overview](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/quality-filtering.md): Compose preprocessing, VAD, band, MOS scoring, and speaker separation stages to extract clean single-speaker training segments from raw audio - [Preprocessing Stages](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/quality-filtering/preprocessing.md): Audio preprocessing stages — mono conversion, segment concatenation, and timestamp mapping for downstream filtering pipelines - [VAD Segmentation](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/quality-filtering/vad.md): Voice activity detection with VADSegmentationStage to split audio into speech segments using Silero VAD - [Band Filter](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/quality-filtering/band-filter.md): Filter audio by spectral bandwidth — full-band vs narrow-band — using BandFilterStage and a scikit-learn classifier - [UTMOS Filter](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/quality-filtering/utmos.md): Filter audio segments by predicted Mean Opinion Score using UTMOSFilterStage and the utmos22_strong model - [SIGMOS Filter](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/quality-filtering/sigmos.md): Filter audio across seven independent perceptual quality dimensions — noise, overall, signal, coloration, discontinuity, loudness, reverb — using SIGMOSFilterStage - [Speaker Separation](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/quality-filtering/speaker-separation.md): Diarize audio with offline SortFormer or streaming Sortformer and fan out one task per speaker for downstream per-speaker filtering - [AudioDataFilterStage Composite](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/quality-filtering/audio-data-filter-stage.md): AudioDataFilterStage composite — end-to-end audio curation pipeline that decomposes into VAD, band, UTMOS, SIGMOS, and speaker separation sub-stages from a single YAML config - [Overview](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/audio-analysis.md): Extract and analyze audio file characteristics including duration calculation, format validation, and metadata extraction - [Duration Calculation](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/audio-analysis/duration-calculation.md): Calculate precise audio duration using soundfile library for quality assessment and metadata generation - [Format Validation](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/audio-analysis/format-validation.md): Audio format support and error handling in NeMo Curator audio processing stages - [Overview](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/alm.md): ALM data curation stages for constructing and filtering training windows from diarized audio segments - [Data Builder](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/alm/data-builder.md): ALMDataBuilderStage reference for constructing training windows from diarized audio segments with quality filtering - [Overlap Filtering](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/alm/overlap-filtering.md): ALMDataOverlapStage reference for removing overlapping training windows based on configurable thresholds - [Text Integration](https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/text-integration.md): Convert processed audio data to DocumentBatch format for downstream processing > **Note:** This page contains both a page directory (above) and the landing page content (below). The page directory is generated for agent use and does not appear on the landing page. > For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.nvidia.com/nemo/curator/v26.04/curate-audio/process-data/llms.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/nemo/curator/_mcp/server. > Process audio data using ASR inference, quality assessment, audio analysis, and text integration for high-quality speech datasets # Process Data for Audio Curation Process audio data you've loaded into `AudioTask` objects using NeMo Curator's comprehensive audio processing capabilities. NeMo Curator provides a specialized suite of tools for processing speech and audio data as part of the AI training pipeline. These tools help you transcribe, analyze, filter, and integrate audio datasets to ensure high-quality input for ASR model training and multimodal applications. ## How it Works NeMo Curator's audio processing capabilities are organized into five main categories: 1. **ASR Inference**: Transcribe audio using NeMo Framework's pretrained ASR models 2. **Quality Assessment**: Calculate and filter based on transcription accuracy metrics 3. **Quality Filtering**: Segment, filter, and diarize raw audio into clean single-speaker training segments 4. **Audio Analysis**: Extract audio characteristics like duration and validate formats 5. **Text Integration**: Convert processed audio data to text processing workflows Each category provides GPU-accelerated implementations optimized for different speech curation needs. The result is a cleaned and filtered audio dataset with high-quality transcriptions ready for model training. --- ## ASR Inference Transcribe audio files using NeMo Framework's state-of-the-art ASR models with GPU acceleration. #### [NeMo ASR Models](/curate-audio/process-data/asr-inference/nemo-models) Use pretrained NeMo ASR models for accurate speech recognition pretrained multilingual gpu-accelerated #### [Batch Processing](/curate-audio/process-data/asr-inference) Efficiently process large audio datasets with configurable batch sizes batch-inference memory-optimization scalable ## Quality Assessment Evaluate and filter audio quality using transcription accuracy and audio characteristics. #### [WER Filtering](/curate-audio/process-data/quality-assessment/wer-filtering) Filter audio samples based on Word Error Rate thresholds accuracy quality-metrics filtering #### [Duration Filtering](/curate-audio/process-data/quality-assessment/duration-filtering) Filter audio samples by duration ranges and speech rate metrics duration speech-rate range-filtering ## Quality Filtering Compose VAD, band, UTMOS, SIGMOS, and speaker-separation stages to extract clean single-speaker training segments from raw audio. #### [Quality Filtering Overview](/curate-audio/process-data/quality-filtering) End-to-end pipeline of preprocessing, segmentation, and filtering stages vad mos-scoring diarization #### [AudioDataFilterStage Composite](/curate-audio/process-data/quality-filtering/audio-data-filter-stage) Single composite stage that decomposes into the full filtering pipeline from a YAML config composite yaml-config end-to-end ## Audio Analysis Extract and analyze audio file characteristics for quality control and metadata generation. #### [Duration Calculation](/curate-audio/process-data/audio-analysis/duration-calculation) Calculate precise audio duration using soundfile library soundfile precision metadata #### [Format Validation](/curate-audio/process-data/audio-analysis/format-validation) Validate audio file formats and detect corrupted files validation error-handling format-support ## ALM Data Curation Curate training data for audio language models by extracting fixed-duration windows from diarized audio segments. #### [ALM Data Builder](/curate-audio/process-data/alm/data-builder) Construct candidate training windows from consecutive segments with quality filtering windowing speaker-count bandwidth #### [ALM Overlap Filtering](/curate-audio/process-data/alm/overlap-filtering) Remove redundant overlapping windows based on configurable thresholds deduplication overlap-ratio target-duration ## Text Integration Convert processed audio data to text processing workflows for multimodal applications. #### [Audio-to-Text Conversion](/curate-audio/process-data/text-integration) Convert AudioTask objects to DocumentBatch for text processing format-conversion pipeline-integration multimodal