voice100-runtime 1.6.1

Voice100 Runtime is a TTS/ASR sample app that uses ONNX Runtime, WORLD and Voice100 neural TTS/ASR models on Python. Inference of Voice100 is low cost as its models are tiny and only depend on CNN without recursion.

Voice100 Runtime

Voice100 Android App is a TTS/ASR sample app that uses ONNX Runtime, WORLD and Voice100 neural TTS/ASR models on Python. Inference of Voice100 is low cost as its models are tiny and only depend on CNN without recursion.

• Beginnings are apt to be determinative and when reinforced by continuous applications of similar influence. Sample audio
• mata toojinoyoonigodaimyooootoyobarerushuyoonamyoooonochuuoonihaisarerukotomoooi Japanese sample audio

Install

Use pip to install from GitHub.

pip install git+https://github.com/kaiidams/voice100-runtime.git

List available models

import voice100_runtime
print(voice100_runtime.list_models())

Using TTS

This downloads ONNX files in ~/.cache/voice100_runtime/ if not available.

import soundfile as sf
import voice100_runtime
tts = voice100_runtime.load("tts_en")
waveform, sample_rate = tts("Hello, world!")
sf.write("output.wav", waveform, sample_rate, "PCM_16")

Using ASR

This downloads an ONNX file in ~/.cache/voice100_runtime/ if not available.

import soundfile as sf
import voice100_runtime
asr = voice100_runtime.load("asr_en")
waveform, sample_rate = sf.read("output.wav")
text = asr(waveform, sample_rate)
print(text)

Using multi-task TTS

This downloads ONNX files in ~/.cache/voice100_runtime/ if not available.

import soundfile as sf
import voice100_runtime
tts = voice100_runtime.load("tts_en_mt")
waveform, sample_rate, align = tts("Hello, world!", return_align=True)
print("/".join(align))
sf.write("output.wav", waveform, sample_rate, "PCM_16")