Uneven volume between podcast episodes, or between voice and background music in a video, is a common technical problem. AI auto-loudness normalization handles this automatically — but there are limitations to understand before publishing.
What is loudness normalization?
Loudness normalization brings audio/video volume to a standard target level. Common standards:
- -14 LUFS: YouTube and Spotify standard
- -16 LUFS: Apple Podcasts standard
- -23 LUFS: Broadcast standard
LUFS (Loudness Units relative to Full Scale) measures perceived loudness as heard by human ears — different from simple dB levels.
Popular AI loudness normalization tools
Adobe Podcast (Enhance Speech)
Adobe Podcast's "Enhance Speech" feature automatically normalizes loudness and reduces noise. Upload a file, AI processes it in seconds. Best for voice recordings — not for music or complex audio.
Auphonic
Auphonic specializes in audio post-production with loudness normalization to platform standards (choose target: YouTube, Apple Podcasts, Spotify, etc.). Free tier available with monthly processing hour limits.
Descript
Descript has loudness normalization integrated into its podcast editing workflow. If you're already using Descript for editing, this is the most convenient option.
DaVinci Resolve (Fairlight)
Free and powerful, but requires learning. Fairlight inside DaVinci Resolve has a loudness meter and normalization tool — ideal if you're already editing video in DaVinci.
Practical workflow
- Export raw audio from your video editor
- Upload to Auphonic or Adobe Podcast — select target loudness (-14 LUFS for YouTube, -16 LUFS for podcast)
- AI processes and exports the normalized file
- Mandatory listen test: Listen back in full (or skip through sections) to check for artifacts, over-compression, or sections the AI processed incorrectly
- Re-import into your video editor if needed
IMPORTANT — AI output is a draft, not a final: AI loudness normalization can create artifacts (strange sounds), over-compression (audio sounds "squashed" and unnatural), or mishandle music/SFX sections. You MUST listen-test after AI processing — never publish AI output directly without reviewing it.
When AI normalization isn't enough
- Files with multiple speakers at very different volumes — manual gain staging needed first
- Files with heavy background noise — noise reduction needed before normalization
- Music and voice require manual balancing — AI cannot interpret your creative intent
Also see AI noise reduction for podcast and voiceover and AI silence removal for video.
Download sample podcast episodes to study audio techniques via Klypio podcast downloader or send a link to @KlypioBot.