How to mash up two songs with AI
Start by deciding what you want to preserve. An AI fusion and a precise vocal-over-instrumental edit require different workflows.
1. Decide what the finished track should do
If you want the exact vocal from one song over the instrumental from another, work with separated stems in an audio editor. If you want a new interpretation inspired by two references, an AI song mashup is the more relevant workflow. Read the AI fusion versus stem mixing comparison before spending credits.
2. Choose two recordings you can use
Upload two MP3 or WAV files, each up to 20 MB. The provider can reject recordings that match its catalog, including licensed recordings; permission alone does not guarantee acceptance. A clearly audible melody and an identifiable rhythmic character make the references easier to describe. Very quiet recordings, heavy clipping and long unrelated introductions can make it harder to assess what the generation should borrow.
A useful pairing might be your own piano demo and your own drum-and-bass loop. Write down the part you like in each before generating: the piano’s calm mood, the loop’s energy, or the contrast between them.
3. Describe the combination, not just the genre
Try “gentle piano harmony, energetic breakbeats, spacious instrumental arrangement” instead of “good mashup.” Use the style box for musical direction; put any original lyrics in the separate optional lyrics field. Your instructions guide the generation but do not guarantee exact melody, tempo or vocal identity.
4. Choose a useful target length
The tool accepts a requested length of 10–360 seconds. The provider may return a longer or shorter track, so this setting does not guarantee an exact duration. Each generation still uses one credit. Use an audio editor afterward if your project needs a precise running time.
5. Listen for coherence before novelty
Our real AI mashup example includes both inputs and both generated takes. It also shows why an exact output length should not be assumed.
Evaluate three things: whether the new arrangement feels musically coherent, whether it expresses the characteristics you wanted from both references, and whether the audio is suitable for your intended use. A surprising result is not necessarily a useful one. Change one part of your style description at a time when making another paid request.
6. Download the result and keep the source rights
Completed audio appears in your private account library. Download the version you want to keep. Generation credits pay for processing; they do not clear copyrights in uploaded recordings or guarantee that a result is suitable for commercial release.