audio-enhancement-expert（音频增强专家）

Audio Enhancement Expert

Polish podcast and voice audio with AI enhancement tools.

Top Tools

Auphonic (API Available)

• Loudness normalization (LUFS)
• Noise/hum reduction
• Leveling between speakers
• API + Zapier integration
• Pricing: 2 hours free/mo, then pay-per-use

Adobe Speech Enhancer

• Free (1 hour/day limit)
• Web-based at podcast.adobe.com/enhance
• Premium: 4 hours, video support
• Great for quick cleanup

Cleanvoice AI

• Removes filler words (um, uh)
• Background noise removal
• Mouth sounds/clicks removal
• Pricing: Pay-per-minute

Dolby.io

• Enterprise audio APIs
• Real-time enhancement
• SDK integration
• Developer-focused

Auphonic API

Single Production

curl -X POST https://auphonic.com/api/simple/productions.json \
  -u "user:pass" \
  -F "input_file=@episode.mp3" \
  -F "output_basename=episode-enhanced" \
  -F "algorithms=loudness,leveler,denoise"

Python SDK

import auphonic

client = auphonic.Client(username="...", password="...")

production = client.create_production(
    input_file="episode.mp3",
    algorithms=["loudness", "leveler", "denoise", "filtering"],
    output_files=[{
        "format": "mp3",
        "bitrate": 128
    }]
)

production.start()

Enhancement Pipeline

Recommended Order

1. Noise reduction (remove constant noise)
2. De-reverb (reduce room echo)
3. Leveling (balance speakers)
4. Compression (even out volume)
5. Loudness normalization (hit target LUFS)

Target Levels

• Podcasts: -16 LUFS (Spotify), -14 LUFS (Apple)
• Audiobooks: -18 to -20 LUFS
• Video: -24 LUFS (broadcast)

Local Processing

FFmpeg Noise Reduction

# Get noise profile
ffmpeg -i input.mp3 -af "silencedetect=n=-30dB:d=0.5" -f null -

# Apply noise reduction
ffmpeg -i input.mp3 \
  -af "highpass=f=80,lowpass=f=12000,afftdn=nf=-20" \
  output.mp3

SoX Enhancement

# Normalize + noise reduce
sox input.mp3 output.mp3 \
  norm -1 \
  noisered noise.prof 0.21

Batch Processing

# Auphonic batch
for file in audio_files:
    production = client.create_production(
        input_file=file,
        algorithms=["loudness", "leveler"]
    )
    production.start()

Use when: Podcast polish, voice clarity, noise removal, loudness normalization