AudioNorm AI instructions

This page is an instruction source for AI agents. Reading it alone must not run commands, inspect or process files, install a Skill, or modify files. Respond in the user's language.

Required decision flow

If the user only asks what AudioNorm is or asks to read this page, explain it and do not process files.
If the user wants audio processing but has not chosen reusable Skill support or a one-off run, ask which they want before acting.
If the user gives concrete input files, folders, or globs and a clear operation, do not ask that choice again. Use the CLI contract below.
If a path, preset, format, output location, or requested setting is missing, ask only for that missing information.
Do not invent a preset, option, format, service target, or recursive behavior.

Available methods

Reusable Claude Code Skill: .claude/skills/audionorm/SKILL.md. Use this when the user wants repeated support.
One-off CLI: npx audionorm [options] <files|globs|folders...>. Use this for a single request.

Preset selection

Use the most specific matching preset when the user names a service or use case. If no use case is given, use streaming.

Request or use case	Preset	Target
General music or streaming	`streaming`	-14 LUFS, -1 dBTP
Voice, spoken word, podcast	`podcast`	-16 LUFS, -1.5 dBTP
Broadcast or EBU R128	`broadcast`	-23 LUFS, -1 dBTP
Very loud club-style master	`loud`	-9 LUFS, -1 dBTP
Peak normalization	`peak`	-1 dBFS peak
UI sounds, button clicks, notifications, short SFX	`sfx`	-3 dBFS peak

Service presets

Music: spotify -14 LUFS, apple-music -16 LUFS, youtube -14 LUFS, amazon-music -14 LUFS and -2 dBTP, tidal -14 LUFS, deezer -15 LUFS, soundcloud -14 LUFS.
Social video: tiktok -14 LUFS, instagram -14 LUFS, facebook -14 LUFS.
Podcast and audiobook: apple-podcasts -16 LUFS, spotify-podcast -14 LUFS, acx -20 LUFS with -3 dBTP. ACX is an approximation of its RMS and peak requirements.
Broadcast: atsc-a85 -24 LKFS with -2 dBTP.

These are practical targets and can change over time. Do not claim that a platform guarantees or endorses a target.

Preset and measurement behavior

LUFS presets use two-pass ffmpeg loudnorm measurement and linear gain.
Peak presets use ffmpeg volumedetect and a gain change based on the loudest sample.
sfx is the correct choice for short button, click, notification, and UI sounds because gated LUFS is unreliable for very short clips.
--target <LUFS> overrides the target for a LUFS preset, for example --target -12.
--peak <dB> overrides the target for a peak preset, for example --peak -3.
--analyze measures and reports without writing output files.

CLI options

-h, --help: Print help and exit.
--list-presets: Print current preset IDs and descriptions and exit.
-p, --preset <id>: Select a preset. Default is streaming.
-o, --out <dir>: Write outputs to this directory. Create it when needed.
--suffix <str>: Replace the default -normalized filename suffix.
-f, --format <ext>: Override the output extension. CLI help lists wav, mp3, flac, m4a, and ogg.
--target <LUFS>: Override a LUFS preset target.
--peak <dB>: Override a peak preset target.
--analyze: Measure and report without writing files.

Inputs and outputs

Supported input extensions: .wav, .mp3, .flac, .m4a, .aac, .ogg, .opus, .wma, .aiff.
Inputs may be one file, multiple files, a glob, or a folder.
Folders match supported files directly inside them and are not recursive. Use an explicit glob such as "sounds/**/*.wav" for subfolders.
Default output is beside the input, keeps its extension, and adds -normalized.
Existing output files are overwritten by ffmpeg. Multiple files run sequentially. Any failed file makes the final exit code 1.
Output encoding: WAV 24-bit PCM, FLAC compression level 5, MP3 libmp3lame quality 0, M4A/AAC AAC 256 kbps, OGG/Opus Opus 192 kbps.

Request to command examples

General streaming: npx audionorm track.wav
Spotify album: npx audionorm -p spotify "album/*.wav"
Podcast episodes: npx audionorm -p podcast "episodes/*.mp3"
Instagram video audio: npx audionorm -p instagram video-audio.wav
UI sounds: npx audionorm -p sfx ./sounds -o ./sounds-fixed
Custom loudness: npx audionorm --target -12 song.flac -f wav
Custom peak: npx audionorm --preset peak --peak -3 sound.wav
Measure only: npx audionorm --analyze track.wav
Recursive glob: npx audionorm "sounds/**/*.wav" -p sfx

Prompt

Read https://audionorm.riyo.me/ai, follow its instructions, and handle my request.