Implementation:Intel Ipex llm NPU Speech Paraformer

Knowledge Sources	Intel IPEX-LLM
Domains	Speech_To_Text, NPU, Audio
Last Updated	2026-02-09 04:00 GMT

Overview

Concrete tool for speech-to-text transcription on Intel NPU using the Paraformer-Large ASR model with IPEX-LLM's FunAsr integration.

Description

This script loads a Paraformer-Large speech recognition model using IPEX-LLM's FunAsrAutoModel for NPU-accelerated inference. It accepts audio file paths and generates text transcriptions with configurable quantization (default sym_int8) and batch processing parameters. The FunAsr integration provides automatic audio preprocessing and decoding.

Usage

Use this when performing speech-to-text transcription on Intel NPU hardware. The FunAsrAutoModel provides a simplified interface for ASR models that handles audio preprocessing, feature extraction, and decoding internally.

Code Reference

Source Location

• Repository: Intel IPEX-LLM
• File: python/llm/example/NPU/HF-Transformers-AutoModels/Multimodal/speech_paraformer-large.py
• Lines: 1-60

Signature

# Script-based execution with argparse
# Key API:
from ipex_llm.transformers.npu_model import FunAsrAutoModel as AutoModel

model = AutoModel(
    model=model_path,
    load_in_low_bit=args.low_bit,
)
result = model.generate(
    input=args.input,
    batch_size_s=args.batch_size,
)

Import

from ipex_llm.transformers.npu_model import FunAsrAutoModel as AutoModel

I/O Contract

Inputs

Name	Type	Required	Description
repo-id-or-model-path	str	Yes	Path to Paraformer ASR model
input	str	Yes	Audio file path for transcription
low-bit	str	No	Quantization type (default: sym_int8)
batch-size	int	No	Batch size for processing

Outputs

Name	Type	Description
Transcription	Console	Text transcription of audio input
Timing	Console	Inference latency

Usage Examples

Speech-to-Text on NPU

python speech_paraformer-large.py \
    --repo-id-or-model-path "iic/speech_paraformer-large" \
    --input "./audio/sample.wav" \
    --low-bit "sym_int8"

Related Pages

• Environment:Intel_Ipex_llm_NPU_Environment