Speech Recognition Model

Model assetKnowledge & DataKnowledge & Dataarc:SpeechRecognitionModel

A transformer ASR model that encodes audio spectrograms and decodes text with word- or sentence-level timestamp alignment, offered in size tiers trading accuracy for latency.

Responsibility. Converts speech audio to timestamped text.

Also known as: Acoustic model

Direct neighbourhood (hover for relationship types)

Relationships

deployed on structural

Automatic Speech Recognizer abstract Ch7.5
Speech Transcriber Ch2.7

Design guidance

SHOULD select base or small size tiers for production multimodal RAG to balance accuracy and latency.

Quantitative guidance

As stated by the sources; verify before use.

Trained on 680,000 hours of multilingual audio; supports 99 languages; five sizes (Ch2.7).
Tiny 39M parameters near real time on CPU; large 1.5B near-human quality at 0.5x real time on GPU (Ch2.7).
Base/small: 90%+ accuracy on conversational audio; base (74M) transcribes a 60-minute meeting in under 2 minutes on one GPU (Ch2.7).

Classification

Technologies: OpenAI WhisperNVIDIA Riva ASR
Quality attributes: Reliability (ISO/IEC 25010 | NIST AI RMF: valid and reliable)

Sources

Ch2.7: T. Nguyen, "Multimodal RAG Approaches," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 2.7. ISBN: 9798244538229.
Ch7.5: T. Nguyen, "NeMo Curator, Riva Speech AI & Multimodal Integration," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 7.5. ISBN: 9798244538229.