MidiMusic

You start in the song creator and choose between two modes. Simple mode takes a description of the song, such as "a slow indie ballad about leaving home, soft piano, warm female vocals", and the model writes the lyrics for you. Custom mode is for people who already have words: you paste your lyrics with section tags like [Verse], [Chorus] and [Bridge], give the song a title and a short style line, and can set a target length from 10 seconds to 6 minutes or leave it open.

Next you pick a model. Standard is the default with natural vocals and rich detail, Fast is a lighter version of Standard that finishes sooner, Wild is bolder and more experimental, and Long song is a separate model with its own sound. A Cadence option writes lyrics from your description and then sings them. Toggling Instrumental removes the voice entirely, which suits video backgrounds, podcast intros and game loops.

Finished songs land in your library. From there you play both takes, download the one you like, or keep working on it without re-uploading: extend it from a chosen point, restyle it, separate the vocals, split stems, export a WAV or make an MP4 video. If a generation fails on the service side, its credits return to your balance automatically.

• •Text and lyrics to song - Generate a full song from a plain-language description or from lyrics you wrote yourself, with vocals or as an instrumental. Genre, tempo, instruments and voice are set with free-text style words rather than a fixed menu, and most runs return two versions to compare.
• •Choice of models - Standard, Fast, Wild and Long song each have their own character, and the site suggests trying the same prompt on two of them. Standard, Fast and Wild return two songs per run; Long song and Cadence return one.
• •Editing tools for generated songs - Extend a song to add a new section, restyle it in another genre while keeping the melody recognisable, or export an uncompressed WAV copy for mixing and mastering, all from songs already in your library.
• •Vocal remover and stem splitter - Separate any song into an isolated vocal and an instrumental, or break the mix into up to 12 stems or pull out a single instrument. The vocal remover accepts uploaded songs as well as tracks made on the site.
• •Covers, lyrics and video - A song cover generator re-sings an uploaded track in a new style, a lyrics generator and a rap generator draft structured words, Sing in my voice re-sings a vocal line in your own voice from a 5 to 25 second recording (a paid feature limited to your own voice or one you have permission to use), and the music video generator turns a track into an MP4 visualizer.
• •Free browser tools - Tempo and key detection with Camelot codes, audio to MIDI, slowed and reverb edits, MIDI to MP3 and format converters run inside the browser tab, so files never leave your device and no credits are used.

Songwriters can use MidiMusic to hear a lyric sung and try a few arrangements before booking studio time. Content creators and podcasters who need background music, intros or short jingles can generate instrumentals described in a sentence. Producers get stem splitting, vocal removal and audio to MIDI conversion for studying arrangements and pulling ideas into a DAW, while DJs can check tempo and key before a set with the free finder.

Teachers and students can use the stem splitter to show how a song is built one part at a time. Hobbyists with no musical training can start from a description in Simple mode, while people with finished lyrics get more control in Custom mode. Anyone planning to release or monetize songs should note that commercial use depends on the plan active when the song was created: the pricing page lists commercial use on the paid plans, and free-account songs are kept for 30 days rather than 365.