Which AI composer can write music you control, and which one only gives presets?

Before you open any AI composer, write down what the cue actually has to be. Something like this:

Cue 14, "The ferry leaves"
key        D minor, turning to F major for the last eight bars
meter      6/8
tempo      84 BPM
lead       solo cello, harp arpeggios underneath
length     1:30, with a hit point at 0:58 when the horn sounds
delivery   MIDI for the strings, so the orchestrator can revoice them

Now read that spec one line at a time and hold it against the tool you were about to sign up for. Does it have a field for the key? Can you type a tempo, or does it offer "calm" and "energetic"? Does anything leave the tool as notes, or only as a finished recording? Seven lines, and most of the "AI composer" market splits cleanly in two.

On one side are tools that write notes you can change: a key you pick, a tempo you set, instruments you swap, a MIDI file you take into your own software. On the other are tools that hand you a finished recording built from presets: you choose a genre, a mood, a length, an energy curve, and you get audio. Both are useful. They are not the same job, and the marketing for both uses the same word.

This page is for the reader who needs that first kind of control and wants to know which tools actually offer it. If you already know the answer is AIVA and your real problem is that one cue needs a voice and words, AIVA composes instrumentals: what do you use when you need a sung song? is the page for you. This one is about the score itself.

One disclosure first. sunomarket is an independent site. We are not AIVA, LANDR, Google, Soundraw, Beatoven or Suno, and we are not affiliated with any of them. What we sell is a lyrics generator and Suno prompt tools, which matter only near the end of this page, when the cue turns out to be a song.

The two families, and the question that sorts them

The sorting question is simple: what comes out of the tool?

If the answer is MIDI or an editable score, you are dealing with a composition assistant. It proposes notes, and you stay the composer. You can transpose the result, rewrite the middle section, give the melody to a different instrument, and nothing about the tool stops you, because once the notes are in your DAW they are just notes.

If the answer is audio, you are dealing with a music generator. It may let you steer a lot before it renders, and some of them let you rearrange the result afterwards, but what you deliver is a recording. You can cut it, layer it, mix its stems. You cannot reach inside and change the third note of the melody.

Neither family is better. A video editor who needs two minutes of upbeat background music by tonight is badly served by a MIDI tool: they would have to orchestrate, mix and render it themselves. A composer on a film is badly served by a preset generator: the director will ask for the cello line a semitone lower, and a recording cannot do that without sounding like a recording that was pitched down.

The trouble starts when you buy the wrong family because both called themselves a composer.

Tools that write notes you can change

AIVA

AIVA is the tool most people mean when they type "AI composer", and it earns the word more than most. According to its own site, it generates music in more than 250 styles, lets you create your own styles, and accepts an audio or MIDI file as an influence for a new piece.

The control lives in the details its user manual describes. In the chord progression workflow, the settings let you set the key signature, the time signature and the tempo before you generate, and you choose the duration and how many tracks to create in one go. After generation, the editor shows the piece as layers: you can change the instrument of a layer from a drop-down, edit the tempo, insert and delete sections, and open the notes in a piano roll where they can be moved, resized, added and deleted. A "Magic" button regenerates notes when a passage is not working.

On delivery, the pricing page lists MP3 and MIDI downloads on the free and Standard plans, and all file formats, including high-quality WAV, on Pro. That MIDI export is the line working composers care about, because it means the cue leaves AIVA as notes rather than as a baked recording.

What AIVA does not do is sing. Its site describes instrumental composition and says nothing about vocals or lyrics.

The catch is rights, and it depends on the plan. At the time of writing, AIVA's pricing page says copyright is owned by AIVA on the free and Standard plans and by you on Pro, with monetisation ranging from none on free, to limited on Standard, to full on Pro, plus an attribution requirement on free. Track length is capped per plan too. Read that page for your plan before anything you make goes into a film, a game or a release.

LANDR Composer

LANDR Composer sits inside your DAW rather than in a browser. LANDR describes it as an AI chord progression generator that creates MIDI sequences for chords, basslines, melodies and arpeggios, with all the standard Western keys and modes available and modifiers for polyphony and humanisation. It syncs with your DAW, and you drag the MIDI patterns straight into the project. It ships as VST3, AU and AAX on macOS and as VST3 and AAX on Windows, comes with over a hundred preset sounds of its own, and at the time of writing is included in LANDR's Studio subscription.

This is the most "you stay the composer" tool on the list. It does not write a cue. It writes material: a progression in the mode you picked, a bassline that follows it, an arpeggio pattern. Arranging that into ninety seconds with a hit point at 0:58 is your job, which is exactly what some people want and exactly what others do not.

Magenta Studio

Magenta Studio comes from Google's Magenta project and is built on Magenta's open-source tools and models. Version 2.0 runs as a device inside Ableton Live (it needs Live 10.1 Suite or later) and works on MIDI clips in Session View. It has five tools: Continue extends a clip by up to 32 measures, Generate makes a four-bar phrase from nothing, Interpolate builds up to 16 clips between two you already have, Groove adds a drummer's feel to a drum clip, and Drumify writes a drum part from the rhythm of any clip.

It is a sketching tool for people who already work in Live. You will not get a finished cue out of it. You will get a continuation of the four bars you were stuck on, which is sometimes the only thing you needed.

Tools that hand you a finished recording

Soundraw

Soundraw is the strongest case for the preset side, because it gives you more steering than the word "preset" suggests. Its site lists more than 30 genres, which you can blend (hip-hop with orchestra, in its own example), a mood setting, and a mixer where you set the length, customise energy, toggle instruments and adjust tempo, key and instrument layers. Its FAQ also mentions bar-level editing: mute, solo, intensity.

What comes out is audio. Depending on the plan, that is MP3 only, or MP3, WAV and stems. MIDI export is not on its feature list. The licence, per Soundraw's FAQ at the time of writing, is worldwide and perpetual, and a track made while you were subscribed stays licensed after you cancel.

So Soundraw can get you close to a spec like the ferry cue on mood, length and energy, and it advertises key and tempo adjustment. It will not give the orchestrator a MIDI file of the cello line.

Beatoven.ai

Beatoven.ai is built for background music for content. Its homepage asks you to describe the background music that fits your content, mentions multimodal prompts, and offers downloads as MP3 or WAV. The homepage lists no fields for key, tempo or instruments.

The terms are the part to read slowly. Beatoven's FAQ says you get a non-exclusive perpetual licence for tracks you generate and download, Beatoven remains the owner of those tracks, and direct distribution of music made with it on streaming platforms is not allowed. That is a workable deal for a video soundtrack and the wrong deal for a cue you want to release on an album.

Suno

Suno belongs on this list because people search for a composer and land on it. It is a different animal again: you describe a style in words, you can paste lyrics, and it renders a full track, vocals included. Of everything on this page, it is the only tool that sings.

As a composer's tool it gives you less direct control than AIVA. A written key or a numeric tempo in the Style field is not reliably honoured, which is why our guide on controlling tempo in Suno steers with feel words instead of numbers. At the time of writing, Suno's pricing page puts Suno Studio, its browser workstation with MIDI, on the Premier plan, and its help centre describes extracting stems from a clip and choosing Get MIDI for a stem, at 10 credits each time. That is MIDI transcribed from audio after the fact, one stem at a time. It is useful for borrowing a part, and it is not the same thing as composing in notes from the start. Commercial use rights are listed on the paid plans and not on the free one.

The control table

Here is the ferry spec held against each tool, using only what each company says on its own pages at the time of writing. "Not listed" means the feature is absent from the pages we checked, not that it is impossible.

Read across a row, not down a column. If your spec has a "delivery: MIDI" line, three columns survive cleanly, and Suno survives only by transcribing its own audio stem by stem. If the spec says "needs to be sung", one column is left.

Which one fits which reader

You score films or games and hand off to an orchestrator or a mixer. Start with AIVA. It is the only one here that takes key, meter and tempo up front, writes a whole piece, and lets you open the result in a piano roll and export MIDI. Check the rights line for your plan before the first cue ships.

You produce in a DAW and want ideas, not finished pieces. LANDR Composer if you want progressions, basslines and melodies in a chosen mode; Magenta Studio if you live in Ableton Live and want to extend or bridge clips you already started. Both keep every decision with you.

You edit video and need music that fits by tonight. Soundraw or Beatoven.ai. Both give you a recording without asking you to orchestrate anything. Soundraw gives you more steering, and stems on its higher plans. Beatoven is quicker to describe, but its licence does not cover releasing the music on streaming platforms.

You need a song: a voice, a chorus, words. None of the composer tools will do it. That is Suno's job, or a singer's, and the words have to come from somewhere. For the full split between instrumental and sung tools, the best AI music generator by job sorts them by what you hand over at the end. If you need Suno to stay instrumental for a background cue, this guide covers stopping the surprise voice.

Where we fit, and where we do not

If you came here to compose a score, we are not your tool. We do not generate audio, we do not write MIDI, and nothing on sunomarket will put a cello line in D minor into your DAW. Use AIVA, LANDR Composer or Magenta Studio for that.

We fit at one point: the cue that turns into a song. The end credits, the tavern song, the radio track in a car scene. There the missing input is not notes but words, and the lyrics generator writes them. You fill in a short form, and it returns a title and a full lyric with [Verse] and [Chorus] tags already in place, ready to paste into Suno's Lyrics field. It does not sing; the audio is a separate step in the tool you pick. A brief for a credits song looks like this:

theme      End credits of a short film. An old man repaints a boat
           nobody has sailed in twenty years; in the last shot his
           grandson takes the oars for the first time.
POV        narrator
language   en
structure  ballad
hook       title-line
emotion    longing
vocal      male
age        any

narrator keeps the song outside the grandfather's head, which suits credits that look back on the whole film. title-line makes the hook the song's own name, handy when the track gets listed in the credits. The theme hands the generator one image and one turn instead of a mood.

SM1, the base model, costs 20 credits, which is $0.20 at 100 credits to the dollar. SM2 and SM3 run an extra editing pass, where a detector flags weak lines and a repair step rewrites them, and cost 80 credits ($0.80). There is no free trial: the cheapest start is the $5 monthly plan with 500 credits a month (25 SM1 lyrics), and the alternative is $19 once for lifetime access and 1 500 credits (75 SM1 lyrics) that never expire. Plans and packs are on the pricing page.

The other half is the style line. The style catalogue holds nearly 2 000 Suno style recipes, and our prompt generator builds a style string from your choices. Everything there is written for Suno's format. Other tools accept a lyric as plain text too, but each handles section tags its own way.

One more honest limit. Suno writes its own melody, and our generator writes no melody at all. Neither will quote the theme you built in AIVA, however carefully you describe it. If the credits song must carry your leitmotif note for note, the realistic route is a human singer over your own arrangement.

Rights, briefly and without a verdict

Every tool on this page has its own terms, and they differ more than the feature lists do. AIVA ties copyright and monetisation to the plan. Soundraw grants a perpetual licence. Beatoven keeps ownership and grants a non-exclusive licence with a streaming restriction. Suno ties commercial rights to paid plans. For a lyric from us, our disclaimer allows commercial use, claims no authorship, and makes the rights non-exclusive: the same model writes for every user, so passages can overlap with someone else's. Whether AI-assisted music is protectable at all is answered differently in different countries, and who owns AI-generated music walks through the main positions.

This is general information, not legal advice. If a contract asks you to warrant who owns a cue, show the relevant terms to a lawyer in your jurisdiction.

Run the spec test before you pay for anything

Take one real cue from your current project, not a demo idea, and write its spec in the seven lines above: name, key, meter, tempo, lead instrument, length with hit points, delivery format. Then open each candidate's feature page and tick the lines it has a field for. The tool that ticks your delivery line and your key line is your composer. The one that only ticks length and mood is a background music generator, which is fine, as long as that is what you were shopping for.

If the cue that fails every tool is the one with a voice in it, that is the moment to write the lyric and hand it to a tool that sings.