How to Make a Mashup with AI: A Step-by-Step Guide

To make a mashup with AI, take the vocals from one song and the music from another, then make them fit. AI handles the part that used to be impossible at home: splitting a finished song into a clean acapella and a clean instrumental. Lining the two up is something you can do yourself, in a free editor like Audacity, in about half an hour.

This guide walks through the whole job in seven steps. It covers how to pick two songs that suit each other, split them with the acapella extractor and the instrumental maker, match tempo and key, line up the vocal, and export a finished mashup.

Key Takeaways

A classic mashup is the vocal of song A over the instrumental of song B. AI stem separation gives you both parts from songs you already have.

Pick songs whose tempos are close and whose keys sit next to each other on the Camelot wheel. Every change you make later adds artifacts.

Stretch the tempo by a few percent and shift pitch by a semitone or two at most. Audacity's own manual warns that big moves leave audible artifacts.

Line up the first downbeat of the vocal with a downbeat of the instrumental, then check the chorus before you export.

AI splits the two songs. You match, align, and mix the two parts.

What Do You Need to Make a Mashup with AI?

You need two songs, a stem separation tool, and an audio editor with more than one track. Nothing here costs money for a first attempt.

Two songs you own a copy of. Use the best-quality file you have. A WAV or FLAC separates more cleanly than an MP3 that has been compressed several times.
An AI vocal remover. Vocal Isolator splits any song into vocals and instrumental and returns both as 44.1 kHz, 16-bit WAV. You can preview the first 30 seconds free without an account, and your first minute of downloads is free after you sign in.
A multitrack audio editor. Audacity is free on Windows, macOS, and Linux and has every effect this guide uses. GarageBand, Reaper, Ableton Live, and FL Studio work too.
The BPM and key of each song. DJ software like rekordbox or Serato shows both after analysis, and so does Mixed In Key. Song databases online list them for most released tracks.
Time: 30–60 minutes for a first mashup. Difficulty: beginner.

How Do You Pick Two Songs That Mash Up Well?

Pick a vocal song and a backing song whose tempo and key already sit close together. AI can split any song cleanly, but it cannot make two mismatched songs agree. Every later step, from stretching to pitch shifting, costs some quality, so the less you need to change, the better the mashup sounds.

Check four things before you split anything:

Tempo. Look for BPMs within a few percent of each other. A song at half or double the tempo of the other also works: an 85 BPM vocal sits on a 170 BPM beat with no stretching at all.
Key. Use the Camelot wheel, which gives every key a number and a letter. Mixed In Key's guide to the Camelot wheel says a track in 8A mixes well into 8A, 7A, 9A, or 8B: the same code, one step around the wheel, or the same number with the other letter.
A vocal with room around it. Rap verses and clear lead vocals extract cleanly. A choir, a vocal drenched in reverb, or a crowd singing along gives a busier acapella.
A beat that repeats. A backing track built on a steady loop is easier to sing over than one with key changes, breakdowns, and tempo shifts.

Song A key	Easy matches for song B	Why it works
8A (A minor)	8A, 7A, 9A, 8B	Same key, neighbor keys, or the relative major
5B (E♭ major)	5B, 4B, 6B, 5A	Same rule, major side of the wheel
Anything else	Shift one song by 1–2 semitones	Moves it to a neighbor key, with small artifacts

Mixed In Key itself calls the wheel a guide, not a rulebook, so trust your ears over the chart once the two parts are playing together.

Step 1: Extract the Acapella from Song A

By the end of this step you'll have the vocal from song A as its own WAV file, with no drums or bass underneath.

Open the acapella extractor and upload song A. Audio and video files both work, including MP3, WAV, M4A, MP4, and MOV, up to 2 GB.
Listen to the free 30-second preview. Solo the vocal and listen to the gaps between lines: a clean acapella has no hi-hats or snare in the silence.
If the intro has no singing, choose Try another part to preview a verse or the chorus instead.
Download the acapella as WAV. Keep the matching instrumental too, in case you want to swap the roles later.

The acapella extractor runs BS-RoFormer, which scores 11.79 dB vocal SDR on the public MVSEP Multisong leaderboard. Older Demucs models score below 8.4 dB on the same test. In a mashup that gap matters more than usual: the acapella plays over a different song, so any drum bleed from song A clashes with song B's drums instead of hiding under them.

The acapella starts on the same sample as the original and runs the same length. If song A has a steady tempo, the vocal stays on that tempo all the way through, which makes Step 4 a single stretch rather than a series of fixes.

Step 2: Make the Instrumental from Song B

By the end of this step you'll have the music from song B with every vocal removed and the bass and drums intact.

Upload song B to the instrumental maker.
Preview the chorus, where vocals are loudest. The voice should be gone while the bass, kick, and every instrument play on at full level.
Download the instrumental as WAV.

The instrumental maker uses the same BS-RoFormer model, which scores 18.10 dB instrumental SDR on MVSEP. Free tools that use phase cancellation subtract the center of the stereo image, and that takes the bass and kick drum with the voice. A thin instrumental under a full vocal sounds wrong straight away, so a clean, full-range backing track matters here.

If you want song B's backing vocals to stay in, so the mashup has harmonies under the new lead, use the karaoke maker instead. It removes only the lead vocal.

Step 3: Find the BPM and Key of Both Tracks

By the end of this step you'll know how far apart the two tracks are, and how much stretching and pitch shifting they need.

Run both files through your DJ software or key finder, or look the songs up in an online BPM and key database. Write down four numbers: the BPM and key of the acapella, and the BPM and key of the instrumental. A stem keeps the tempo and key of its source song, so the original song's values apply.

Then work out the tempo change with this formula:

Tempo change (%) = (target BPM ÷ current BPM − 1) × 100

To move a 120 BPM vocal onto a 124 BPM beat, the change is (124 ÷ 120 − 1) × 100 = +3.3%. Here is how far a few common pairs need to move:

The smaller the bar, the cleaner the stretched track sounds.

For key, count semitones between the two keys. One semitone is a pitch change of about 5.9%, since each step multiplies the frequency by the twelfth root of two. Shifting a vocal by one or two semitones usually still sounds natural. Beyond that, the voice starts to sound like a different singer.

Step 4: Match the Tempo Without Changing Pitch

By the end of this step both tracks will play at the same BPM.

Open Audacity and import both WAV files with File > Import > Audio. Each file gets its own track.
Decide which track to stretch. Stretch the one that needs the smaller change. If the numbers are close, stretch the instrumental: people notice a warped voice sooner than a warped drum loop.
Select the whole track you're stretching.
Open Effect > Pitch and Tempo > Change Tempo, enter the BPM values or the percentage from Step 3, and apply it.

Audacity's Change Tempo effect changes the length of a selection without changing its pitch. Don't use Change Speed for this step, because it changes the pitch along with the tempo. The current Audacity manual lists Sliding Stretch as the replacement for Change Tempo and Change Speed, so if your version doesn't show Change Tempo, use Sliding Stretch with the same initial and final tempo values.

To check the result, play both tracks together over an 8-bar stretch. If the drums and the vocal phrasing drift apart over that time, the tempo is still slightly off.

Step 5: Match the Key

By the end of this step the vocal will sit in a key that fits the instrumental.

If Step 3 showed the keys already match or sit side by side on the Camelot wheel, skip this step.
Otherwise select the track that needs to move. That is usually the instrumental, for the same reason as in Step 4.
Open Effect > Pitch and Tempo > Change Pitch and shift it by the number of semitones you counted. Change Pitch leaves the tempo alone, so the work from Step 4 stays intact.
Play the chorus of both tracks together and listen for sour notes on long, held vocal lines.

The Audacity manual warns that changing pitch or tempo on its own means resynthesising the audio, so large changes leave audible artifacts. If a song pair needs more than a two-semitone shift and a large tempo change, a different song B will usually sound better than heavier processing.

Step 6: Line Up the Vocal on the Beat

By the end of this step the vocal will land on the beat of the instrumental from the first line onward.

Find the first downbeat of the vocal phrase, usually the first stressed word of a verse or chorus. Zoom in with View > Zoom In until you can see it clearly.
Find a bar start in the instrumental. Analyze > Beat Finder adds a label track that marks beats. The Beat Finder manual page calls it rough and ready, so use the labels as a guide and confirm by ear.
Drag the vocal clip with the Time Shift tool, or the clip's top bar in newer versions, until its downbeat sits on the bar start.
Play the first eight bars. If the vocal sounds early or late, nudge it a few milliseconds at a time.

Aligning a whole verse with a whole section is easier than aligning single words. Cut the acapella into sections, such as verse, pre-chorus, and chorus, and place each one where it fits the instrumental's arrangement. A chorus vocal over the instrumental's chorus usually lands harder than one over a quiet verse section.

Step 7: Mix and Export the Mashup

By the end of this step you'll have one finished file you can share.

Balance the levels. Use each track's gain slider so the vocal sits on top of the beat without shouting over it. Start with the instrumental around 3–6 dB below the vocal and adjust by ear.
Clear space for the voice. Add a gentle cut to the instrumental's midrange with Effect > EQ and Filters > Filter Curve EQ, and roll off the low end of the vocal track below about 100 Hz. A lead vocal has little useful energy that low, and the cut removes any trace of song A's bass that would clash with song B's.
Fade the edges. Add a short fade-in and fade-out with Effect > Fading, so the mashup doesn't start or stop abruptly.
Check for clipping. Play the loudest chorus. If the meter hits the red, lower both tracks slightly instead of only the vocal.
Export. Use File > Export Audio and save a WAV master. Make an MP3 copy if you're uploading to a platform that prefers small files.

Listen once on headphones and once on a phone speaker before you share it. A phone speaker shows quickly whether the vocal still cuts through when the bass disappears.

Why Does My Mashup Sound Wrong?

Most mashup problems come from one of five causes, and each one has a quick check.

What you hear	Likely cause	Fix
The vocal drifts off the beat over time	Tempo is slightly off	Recheck both BPMs and redo Step 4
Long notes sound sour against the chords	Keys clash	Shift one track by a semitone, or pick another song B
A watery or robotic voice	Too much stretching or pitch shifting	Move the instrumental instead, or choose a closer pair
Two drum kits playing at once	Drums bled into the acapella	Use a cleaner source file and preview the gaps before download
The vocal disappears in the chorus	Both parts fight for the same midrange	Cut the instrumental's mids and lower it 1–2 dB

Can You Post a Mashup Online?

A mashup uses two copyrighted recordings, so posting it publicly depends on the rights holders and the platform, not on the tool you used. Platforms such as YouTube scan uploads for copyrighted audio, and a mashup can be claimed, muted, or blocked. To publish one safely, get permission, use songs you hold the rights to, or use tracks released for remixing. For practice, DJ sets you perform live, and private listening, these concerns mostly don't apply.

Is There a One-Click AI Mashup Maker?

Yes. Some apps and online tools blend two songs automatically by matching BPM and key and picking the transition for you. They're fast, but you can't choose which vocal section lands where, and you usually can't fix bleed or a clashing note. The step-by-step method above takes longer, but you control every bar, and you get studio-quality stems you can reuse in remixes, edits, and DJ sets.

Frequently Asked Questions

Can I make a mashup with AI for free?

Yes, for a first attempt. Vocal Isolator previews the first 30 seconds of any split free without an account, and your first minute of downloads is free after you sign in. Audacity is free. Longer songs use minutes from a minute pack or monthly plan, charged by song length.

Which song should give the vocal and which the beat?

Give the vocal role to the song with the clearer, more recognizable vocal, and the beat role to the song with the steadier, more repetitive instrumental. Recognizable hooks over an unexpected beat are what make most mashups land.

Do I need DJ software to make a mashup?

No. A multitrack editor like Audacity is enough to stretch, shift, align, and export. DJ software helps you find the BPM and key, and it lets you perform mashups live, but it isn't required to make a finished file.

Can I use more than two songs?

Yes. Split each extra song the same way and treat each part as another track. Keep every part at the same BPM and in compatible keys. Each added song makes the mix busier, so cut the parts you add back to short sections.

How do I mash up a song from a video?

Upload the video file itself, such as an MP4 or MOV, to the acapella extractor or instrumental maker. The tool separates the audio track the same way it handles a song file.

Make Your First Mashup

A mashup with AI comes down to one split and one fit:

Choose two songs with close tempos and neighboring keys.
Pull the acapella from song A and the instrumental from song B with an AI vocal remover.
Stretch and shift as little as you can, then line up the first downbeat and mix.

Start with the vocal. Upload song A to the acapella extractor, solo the vocal in the free preview, and check how clean it is before you build anything on top of it. If you'd rather start from the beat, the instrumental maker works the same way, and the vocal isolator splits both parts at once.