AidVoice SDK

Introduction

AidVoice SDK is an AI inference SDK specifically designed for voice-related models launched by Aplux. It aims to simplify the development of core voice processing functions based on edge AI technology, allowing for flexible and rapid integration into intelligent applications. The SDK provides a unified and efficient API, supporting industry-leading voice processing AI models to meet the requirements of various business scenarios.

Support Status

Operating System Support

	Linux	Android
C++	✅	／
Python	🚧	／
Java	／	✅

ASR Model Support

Whisper-Base-en	SenseVoiceSmall	Whisper-tiny	Whisper-small	Whisper-medium
✅	✅	✅	✅	✅

TTS Model Support

MeloTTS-Chinese	MeloTTS-English
✅	✅

✅: Supported 🚧: Planned

SDK for Linux

Quick Start

Installation

# Install AidLite QNN version 2.36
sudo aid-pkg update
sudo aid-pkg install aidlite-sdk
sudo aid-pkg install aidlite-qnn236

# Install AidVoice SDK
sudo aid-pkg -i aidvoice-sdk

Model Acquisition

Obtain the corresponding model using the mms tool, taking SenseVoiceSmall as an example:

mms list sensevoicesmall

#------------------------ SenseVoiceSmall models available ------------------------
Model             Precision  Chipset           Backend
-----             ---------  -------           -------
SenseVoiceSmall   W8A16      Qualcomm QCS8550  QNN2.31
SenseVoiceSmall   FP16       Qualcomm QCS8550  QNN2.31
SenseVoiceSmall   W8A16      Qualcomm QCS6490  QNN2.31

# Download SenseVoiceSmall
mms get -m SenseVoiceSmall -p fp16 -c qcs8550 -b qnn2.31 -d /home/aidlux/sensevoicesmall

Development Workflow

ASR

TTS

Examples

For more cases, please refer to the <a href="/software/tutorial" target="_blank">Case Center</a>.

ASR: Recognizing Audio Files on QCS8550 Linux System

Install AidVoice SDK

# Install AidLite QNN version 2.36
sudo aid-pkg update
sudo aid-pkg install aidlite-sdk
sudo aid-pkg install aidlite-qnn236

# Install AidVoice SDK
sudo aid-pkg -i aidvoice-sdk

Model Acquisition

# Obtain Whisper-small

mms get -m Whisper-small -p fp16 -c qcs8550 -b qnn2.31 -d /home/aidlux/whisper-small

cd /home/aidlux/whisper-small
unzip whisper-small_qcs8550_qnn2.31_fp16_aidlite.zip

Test Code Compilation

# Copy test code
cp -r /usr/local/share/aidvoice/examples /home/aidlux/aidvoice

# Compile
cd /home/aidlux/aidvoice/asr/cpp/
mkdir -p build && cd build
cmake ..
make

Run Example

# -m model path   
# -a audio path (optional, uses default if not provided)
./test_asr -m /home/aidlux/whisper-small/model_farm_whisper-small_qcs8550_qnn2.31_fp16_aidlite/models

TTS: Text-to-Speech on QCS8550 Linux System

Install AidVoice SDK

# Install AidLite QNN version 2.36
sudo aid-pkg update
sudo aid-pkg install aidlite-sdk
sudo aid-pkg install aidlite-qnn236

# Install AidVoice SDK
sudo aid-pkg -i aidvoice-sdk

Model Acquisition

# Obtain MeloTTS-English

mms get -m MeloTTS-English -p fp16 -c qcs8550 -b qnn2.31 -d /home/aidlux/melotts-english

cd /home/aidlux/melotts-english
unzip melotts_en_woman_qcs8550_qnn231_fp16_aidlite.zip

Test Code Compilation

# Copy test code
cp -r /usr/local/share/aidvoice/examples /home/aidlux/aidvoice

# Compile
cd /home/aidlux/aidvoice/tts/cpp/
mkdir -p build && cd build
cmake ..
make

Run Example

cd /home/aidlux/aidvoice/tts/cpp/build
./test_tts -m /home/aidlux/melotts-english/model_farm_melotts_en_woman_qcs8550_qnn231_fp16_aidlite/models

The audio file will be generated in the current directory.

SDK for Android (TBD)