midasheng-audio-denoise（米达声音频降噪）

midasheng-audio-denoise

Voice enhancement and noise reduction service powered by advanced backend algorithms.

1. Trigger

Use this skill when the user wants to:

• Remove background noise from an audio file
• Clean up a noisy recording
• Enhance voice clarity
• Preprocess audio for speech recognition

2. API Details

Endpoint: POST https://llmplus.ai.xiaomi.com/dasheng/audio/denoise (multipart/form-data)

Parameters:

• file: The audio file to denoise

Response: Binary audio stream (WAV format)

3. Usage

Basic denoising

curl -X POST "https://llmplus.ai.xiaomi.com/dasheng/audio/denoise" \
  -F "file=@noisy_recording.mp3" \
  -o clean_recording.wav

Script usage

python3 scripts/denoise.py noisy_audio.mp3 -o clean_audio.wav
python3 scripts/denoise.py --queue   # Check queue status

4. Queue Status（排队情况）

查询命令

python3 scripts/denoise.py --queue
# 或直接调 API：
curl -X POST "https://llmplus.ai.xiaomi.com/metrics?path=/dasheng/audio/denoise"

返回字段

• active: 当前活跃请求数
• avg_latency_ms: 平均处理耗时（毫秒）
• 预估等待时长 = active × avg_latency_ms

何时调用

1. IM 即将超时但 denoise 服务还未返回结果时：查排队情况告知用户，请用户稍后来问。
2. 用户稍后询问任务进度但服务仍未返回时：查最新排队情况返回给用户。

状态分级

• 🟢 active=0 或预估等待 <5s → 服务空闲
• 🟡 预估等待 5-30s → 轻微排队
• 🔴 预估等待 >30s → 排队较长，建议稍后重试

5. Supported Audio Formats

Input: mp3, wav, flac, ogg, m4a. Output: WAV.

6. Troubleshooting

• Empty output: Input file may be too short
• API request failed: Verify network connectivity
• Poor results: Works best on speech with background noise