audionorm

Normalize audio files to an optimal, consistent loudness from the command line — so a clip you grabbed from a free sound library doesn't blast out at full volume, and your app's UI / button sounds all sit at the same comfortable level.

Bundles its own ffmpeg, so there's nothing else to install. Just run it with npx.

npx audionorm track.wav

Prefer a UI? There's also a browser version (drag & drop, nothing uploaded): https://audionorm.riyo.me

Why

Loudness is perceptual, not just peak level — two files with the same peak can sound very different in volume. audionorm measures integrated loudness in LUFS (ITU-R BS.1770 / EBU R128, the same standard Spotify, YouTube and broadcasters use) and applies the gain needed to hit your target, while keeping the true peak under a ceiling so it never clips.

For very short clips (button clicks, notifications) gated LUFS is unreliable, so there's a dedicated peak-based sfx preset that just brings every clip to a consistent, not-too-loud peak.

Usage

npx audionorm [options] <files | globs | folders...>

# Single file → -14 LUFS, writes track-normalized.wav next to it
npx audionorm track.wav

# Fix a folder of too-loud UI sounds into a new folder
npx audionorm ./sounds -p sfx -o ./sounds-fixed

# Batch a glob to podcast loudness
npx audionorm -p podcast "episodes/*.mp3"

# Just measure — don't write anything
npx audionorm --analyze track.wav

# Custom target and convert to WAV
npx audionorm --target -12 song.flac -f wav

Presets

General

| id | Target | Best for | | ----------- | ----------- | ------------------------------------------------- | | streaming | −14 LUFS | General use, music (Spotify/YouTube level) | | podcast | −16 LUFS | Voice, spoken word | | broadcast | −23 LUFS | EBU R128 broadcast | | loud | −9 LUFS | Hot master, maximum perceived loudness | | peak | −1 dB peak | Lossless peak normalize | | sfx | −3 dB peak | UI / button sound effects (the too-loud fix) |

Per-service targets (master to a platform's playback loudness):

| id | Target | Service | | ----------------- | ------------------ | ----------------------------- | | spotify | −14 LUFS | Spotify | | apple-music | −16 LUFS | Apple Music (Sound Check) | | youtube | −14 LUFS | YouTube / YouTube Music | | amazon-music | −14 LUFS, −2 dBTP | Amazon Music | | tidal | −14 LUFS | Tidal | | deezer | −15 LUFS | Deezer | | soundcloud | −14 LUFS | SoundCloud | | tiktok | −14 LUFS | TikTok | | instagram | −14 LUFS | Instagram / Reels | | facebook | −14 LUFS | Facebook | | apple-podcasts | −16 LUFS | Apple Podcasts | | spotify-podcast | −14 LUFS | Spotify (podcast) | | acx | −20 LUFS, −3 dBTP | Audible / ACX audiobook | | atsc-a85 | −24 LKFS, −2 dBTP | ATSC A/85 (US TV) |

Run npx audionorm --list-presets for the current list.

Options

-p, --preset <id>     Loudness preset (default: streaming)
    --target <LUFS>   Override the target loudness for LUFS presets
    --peak <dB>       Override the target peak for peak presets
-o, --out <dir>       Output directory (default: alongside each input)
    --suffix <str>    Filename suffix for outputs (default: "-normalized")
-f, --format <ext>    Output format: wav, mp3, flac, m4a, ogg (default: keep input)
    --analyze         Only measure and report — don't write files
    --list-presets    List the available presets and exit
-h, --help            Show help

How it works

LUFS presets use ffmpeg's two-pass loudnorm (linear mode) for accurate, transparent gain, pulling back automatically if it would push the true peak past the ceiling. Peak presets apply a single lossless gain change. Output quality is preserved (24-bit WAV, FLAC, or high-bitrate lossy depending on the target format).

Links

• Web app: https://audionorm.riyo.me
• Source: https://github.com/Riyoway/audionorm

License

MIT