Separate
Use the vocal remover or stem splitter to hear each element of your recording on its own.
a guide for working from audio
How to Make AI Music by Building on Existing Audio
Most guides start with a prompt. This one starts with a recording and shows how to turn it into a finished track step by step.
a guide for working from audio
The audio-first method
Most guides start with a prompt. This one starts with a recording and shows how to turn it into a finished track step by step.
Step one is to understand what you have. Separate the recording into parts, listen to each, and decide what to keep. Step two is to fill the gaps: add a vocal, add an instrumental, or extend a section that ends too soon.
Step three is to make the result portable. Convert melodies to MIDI, download stems, and bring everything into the software you already use to mix and finish.
AI music does not have to begin from nothing. If you have a voice memo, a rough demo or a beat, you already have the most important part: an idea that is yours.
This guide walks through the SeedAudio tools in the order producers usually use them, from isolating parts to adding new layers and exporting files for your own editor.
The toolchain in order
Add and extend
Fill missing layers with Add Vocal or Add Instrumental, and lengthen the track with the extender.
Convert and export
Transcribe key parts to MIDI, fix them in the editor, and download everything for your DAW.
A worked example
Imagine a thirty-second voice memo with a strong chorus melody and nothing else.
- 1
Build the backing
Run the memo through Add Instrumental with a style direction to hear the chorus as a full-band idea.
- 2
Grow the structure
Use the extender to add a verse and a second chorus, choosing continuation points at phrase endings.
- 3
Take it home
Split the result into stems, convert the melody to MIDI, and rebuild the arrangement in your own DAW with the AI version as a reference.
Who this guide is for
Singers with ideas but no band
Turn sung melodies into demos you can share without learning production first.
Producers with unfinished sketches
Use AI tools to fill gaps quickly, then finish by hand.
Teachers and learners
Use separation and transcription to study how songs are built from their parts.
Principles that keep results usable
Keep your original
Always keep the unprocessed source. It is the reference you return to when a later version drifts.
Change one thing per pass
Make a single change, listen, then decide the next one. Big jumps make it hard to know what helped.
Know your rights
Build on your own recordings or licensed material, and review terms before releasing commercially.
Guide questions
Do I need musical training?
No. Separation, extension and adding layers only require you to listen and choose. Theory helps later when editing MIDI.
Which tool should I start with?
Start with separation if your recording is already a mix, or with Add Instrumental if you only have a vocal idea.
Can I finish the song outside SeedAudio?
Yes. Download stems and MIDI and continue in any DAW you already use.
What if I have no recording at all?
Generate a reference with the music generator, then follow the same steps using that file as your source.
Start with the recording you already have
Pick one memo or demo and take it through the first step of the guide today.
Open the workbench