Media Denoise AI

Context: $ARGUMENTS

Quick start

• Default, real-time speech denoise (MIT): scripts/denoise.py denoise --model deepfilternet --in noisy.wav --out clean.wav → Step 3
• Full restoration (denoise + dereverb + bandwidth extension): scripts/denoise.py enhance --in muffled.wav --out clean.wav → Step 4
• Batch a folder of podcast episodes: scripts/denoise.py batch --model deepfilternet --in-dir raw/ --out-dir clean/ → Step 5
• Use ffmpeg's built-in RNNoise (fastest, no new install): scripts/denoise.py denoise --model rnnoise --in noisy.wav --out clean.wav

When to use

• Remove steady background noise (HVAC, fan hum, traffic) from voice recordings
• Clean keyboard clicks, mouse clicks, fridge hum out of video-call recordings
• Restore old / cassette / lo-fi audio (Resemble Enhance handles bandwidth extension)
• Pre-process audio before ASR (media-whisper) — dramatically reduces hallucinations
• Pre-process audio before TTS voice-cloning reference clips (media-tts-ai clone)
• Clean podcast tracks before loudness-normalization