Tap a one-bar melody onto the grid, or start from a preset, and a neural model writes the next bars in the same style. Edit the result, loop it, and export MIDI. Nothing leaves your browser.
How it works
Click notes onto the one-bar grid to sketch your idea, or load a preset and adjust it. Even four or five notes is enough for the model to catch the feel.
Choose how many bars to add and how adventurous it should be, then hit Continue. The AI writes the next steps one at a time in the style of your seed.
Loop the result to hear it, tweak any notes you want to keep, and download the whole phrase as MIDI for your DAW or notation software.
Who uses it
Five moments where a fresh idea for the next bar keeps a session moving.
The situation: You have a strong opening phrase but no idea where the melody should go next.
What they do: Enter the phrase as the seed, generate a few continuations, and keep the one that surprises you in a good way.
The result: You are unblocked with an idea you can shape rather than a blank bar.
The situation: A producer has a beat and chords but the vocal melody is not landing.
What they do: Sketch a two-bar motif, let the model extend it, and sing along to the exported MIDI.
The result: Several usable topline directions in a few minutes instead of an afternoon.
The situation: A composition student wants to see different ways a motif can be answered.
What they do: Feed in one motif and generate a batch of continuations at different adventurousness settings.
The result: A set of concrete examples to analyse and learn from.
The situation: An indie developer has a short leitmotif but needs a longer loop for a level.
What they do: Use the motif as the seed, extend it to four bars, and export the MIDI to arrange with their own instruments.
The result: A longer, coherent loop that still sounds like the original theme.
The situation: You want call-and-response phrases to practise answering by ear.
What they do: Generate a phrase, play it, then try to sing or play a response before checking what the model did next.
The result: A steady supply of fresh phrases to train your ear against.
FAQ
No. The model runs in your browser and generates everything locally. Your melodies stay on your device.
One bar of sixteen steps. You do not have to fill it — a handful of notes gives the model enough to work with, and rests are fine.
Yes. It is one note at a time, like a sung or whistled line. For chords and drums, use the other Melody Sheet tools.
It is the sampling temperature. Lower keeps the continuation close and predictable; higher takes bigger leaps. Regenerate a few times to compare.
The first time you press Continue, the browser downloads the model (about 2MB) and caches it. After that it responds almost immediately.
The seed bar is fully editable. To rework the generated part, download the MIDI and open it in your editor, or clear and reshape the seed and generate again.
The full phrase — your seed plus the continuation — as a single melody track at the tempo you set, ready for any DAW or notation app.
It tends to follow the tonality of your seed, but it can wander, especially at higher adventurousness. Keep the seed clearly in one key for the steadiest results.
Yes, in a modern mobile browser. The grid is scrollable and playback is light; only the one-time model download needs a connection.
Yes, with no account and no limit on how many melodies you generate or export.