Understand sound. Generate sound.

A free music model that writes whole songs with sung lyrics in more than 50 languages, 10 seconds to 10 minutes long, on your own computer.

Meta's free research models: MusicGen makes music from a description or a tune you give it, and AudioGen makes sound effects.

Writes full songs or instrumentals from a written description, 3 seconds to 5 minutes long, with or without singing.

Turns a short written description into a sound effect or a seamless loop, up to 30 seconds long.

Google's music models: 30-second clips, full songs with singing from words or a picture, and live music you steer as it plays.

Makes royalty-free music from a description, a picture, or a chosen tempo and genre, and can stream endless music live.

Makes instrumental music, single-instrument parts, samples and sound effects from words or from a sound you give it, up to about six minutes.

Meta's free AI model that scores any speech, music or sound clip on four things: how well made, how complex, how enjoyable and how useful.

Amazon's cloud service for converting video and audio files, which can also measure each file's loudness and loudest peak and save the readings.

Software for large media companies that checks video and audio files, including loudness rules, before they go to TV or streaming services.

Free analysis code from a Barcelona university lab: it finds clipping, clicks, gaps and hum, measures loudness and, with add-ons, labels music.

Free software that measures loudness to the European broadcast standard, finds silent stretches and reports how high the peaks go.

Free code for measuring audio, such as its level, silent stretches, where notes start and what pitch is playing, to build your own checks.

Free code that measures how loud a recording sounds, by the international standard, and turns it up or down to a target loudness.

Splits a song or a film soundtrack into its parts, such as vocals, drums, bass, guitar, dialogue and effects, and cleans up speech.

Automatic finishing for spoken audio: evens out voices and music, removes noise, hits a set loudness, and cuts filler words and coughs.

Free, open software that splits a song into vocals, drums, bass and everything else; a six-part version also separates guitar and piano.

Splits songs into separate parts, removes background noise or music from speech, and turns one voice into another.

Automatic mastering: send a finished song and get back a polished, release-ready version, with a choice of three intensities and styles.

Audio tools from the makers of Moises, chained together: split songs into parts, master, mix, add effects, write out chords, beats and lyrics.

Spotify's free code library that adds studio effects such as reverb, delay and compression to audio, and runs music plug-ins on a server.

Recognises which song is playing from a recording or a file, and links it to streaming services and the music industry's track codes.

Spotify's free tool that listens to any instrument or voice and writes down the notes, chords and bends it hears, for music apps.

Listens to music tracks, labels their genre, mood, instruments and more, and finds songs that sound alike in a catalogue.

Searches a large public library of sounds shared under Creative Commons licences, by words, tags or by how they sound.

A free AI model that links sounds and words: search sounds by description, label them with your own words, or check a sound matches its prompt.

Apple's kit that adds Shazam song recognition to apps, and can also recognise your own pre-recorded audio from a catalogue you build.

Google's free model that recognises everyday sounds, such as speech, dog barks, sirens or music, across 521 categories.