fasr-enhancement-hush
英文文档地址: README_EN.md
Hush 音频增强插件。该插件使用 Hush 上游提供的 ONNX 模型包
advanced_dfnet16k_model_best_onnx.tar.gz 和 libweya_nc native runtime,
通过 fasr EnhancementModel 暴露离线 enhance() 与流式 push_chunk()。
注册模型
| 注册名 | 类 | 适用场景 |
|---|---|---|
hush |
HushEnhancement |
16 kHz 语音降噪 / 背景说话人抑制 |
使用方式
from fasr.data import Waveform
from fasr_enhancement_hush import HushEnhancement
model = HushEnhancement(sample_rate=16000)
enhanced = model.enhance(Waveform(data=audio, sample_rate=16000))
Confection 配置
[enhancement_model]
@enhancement_models = "hush"
sample_rate = 16000
attenuation_limit_db = 100.0
参数
| 参数 | 类型 | 默认值 | 说明 |
|---|---|---|---|
sample_rate |
int |
16000 |
输出采样率;Hush 内部原生 16 kHz,其他输入采样率会先重采样 |
model_path |
str | None |
None |
可选 Hush ONNX tar.gz 模型包路径;为空时使用内置模型 |
library_path |
str | None |
None |
可选 libweya_nc / weya_nc.dll 路径;为空时按平台使用内置库 |
attenuation_limit_db |
float |
100.0 |
最大衰减限制;值越大抑制越强,较小值会保留更多背景 |
Realtime 状态
插件实例只加载一次 Hush native model。每个 stream_id 会创建独立 native
session,并在 is_last=True 后释放 session 和清理状态,因此可以配合 fasr
realtime 服务的单模型实例并发使用。
Metadata
Release files for fasr-enhancement-hush 0.5.7
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| fasr_enhancement_hush-0.5.7-py3-none-any.whl | Python 3 | none | any | Details |
Release files / fasr_enhancement_hush-0.5.7-py3-none-any.whl
| Download URL | fasr_enhancement_hush-0.5.7-py3-none-any.whl |
|---|---|
| Size | 46.9 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
1e57d1de9796060d18d47d61b65b378d30cca7b5e4c4ac7887b9098095128421
|
|
BLAKE2b-256 checksum How to use checksums |
afd8ea985267779c38615d3c0fe58a3b5962683f540249fd2a04346041f10a49
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.13.12
|