Music stems are rendered audio files containing separate elements or groups from a song, such as vocals, drums, bass and other instruments. They let you work with parts independently. Before asking for stems, say whether you need grouped files, individual production tracks or parts extracted from an existing mix.
That distinction matters because “send me the stems” can mean different things to different people. A grouped drum file is useful for changing the drum balance, but it does not give the same control as separate kick, snare and cymbal recordings.
Understand the three kinds of material
Grouped stems combine related tracks. A drum stem might contain several percussion instruments, and a backing-vocal stem might combine harmonies. You manage fewer files, while retaining control over broad musical groups.
Individual tracks or parts keep more elements separate. If the next producer needs to change the kick independently from the snare, request those parts explicitly. Ask what each file contains instead of relying on the word “stems.”
AI-separated parts are estimates extracted from an existing mixed recording. They can help when original production tracks are unavailable. They do not necessarily recover those originals exactly, and the source recording and separation method affect the result.
Match the files to the edit
Suppose a collaborator wants to make the chorus vocal clearer. They might need a vocal stem and an instrumental stem for a broad balance change. If a particular synth competes with the vocal, separating that synth from the other instruments would give them more targeted control.
For a detailed remix, the useful request might be individual drums, bass, lead vocal, backing vocals and other instruments. For a simple performance backing track, the request might be the full instrumental plus selected vocal elements. The right file set depends on the actual task.
These are illustrative handoff scenarios. Before exporting everything, send a short list of intended edits to the recipient and agree on the parts that make those edits possible.
Separation and generation solve different problems
Separation starts with existing audio and attempts to isolate components already in it. Generation creates new material. Isolating a bass from a mix and creating a new bass idea are different jobs, even if a product offers both in one workspace.
Mozart lists stem splitting and generation among its Studio features. Check its current product and plan details before choosing a workflow. First name the task: isolate the existing vocal, reduce a competing part, or create something different. That choice determines what you should inspect next.
current product and plan details
Watch a stem-splitting demonstration
Mozart's recorded Studio Fundamentals tutorial reaches a stem-splitting example at about 5:52. It shows a clip-based workflow and the separated parts in the session. Use the video to understand the sequence, while checking today's interface before following an exact control or menu.
A demonstration explains how a task is approached; your own source material still needs evaluation. A sparse recording and a busy mix can expose different problems, so judge the result where you intend to use it.
Listen in context before making a decision
Choose a short section containing the part you care about. Listen to the full mix, then the isolated part, then that part alongside the other material. Keep the playback level comfortable and reasonably consistent so loudness does not dominate the comparison.
Write down what you hear rather than immediately naming an effect. For example: the vocal is quiet, another instrument masks a phrase, or unwanted sound becomes noticeable when the vocal is raised. Those observations point to different next steps.
A balance issue may need a level change. An arrangement issue may need less competing activity. A separation artifact may call for another source, another method or a less exposed use of the extracted part. This is a suggested listening process, not a report of tested Mozart outputs.
Prepare a handoff another person can use
Agree on these details before exporting:
- Parts: name each group or individual track the recipient expects.
- Timing: use an agreed common start and render range so files align.
- Effects: specify whether each file includes processing and effect tails.
- Format: agree on the audio format and any technical delivery requirements.
- Context: include a reference mix and the tempo information you have verified.
- Purpose: explain the requested changes and anything that should stay intact.
Clear filenames help. “Lead-vocal-with-reverb” and “Lead-vocal-dry” communicate more than “audio-final-2,” provided the labels accurately describe the contents. Include a brief note if the effect is printed into the file and cannot be removed independently.
Import the files into an empty session when possible and check their alignment against the reference. That simple handoff check can catch a wrong render range or missing part before your collaborator begins.
Start with one useful task
You do not always need the largest possible collection of files. Begin with the edit you want and the minimum useful separation. Keep the original mix as a reference, and label any extracted material so a collaborator understands how it was obtained.
To explore Mozart, bring one clear task, such as working on an existing vocal or developing a different bass idea. Check the relevant controls and delivery options before expanding the project.
Sources and further reading
Keep exploring
AI music generator or DAW: which do you need?
Understand when to use an AI music generator, a DAW, or both—and the editing and export questions to ask before choosing your workflow.
Read guide : AI music generator or DAW: which do you need?How to write an AI music prompt: a practical brief and three examples
Build a clear AI music prompt with style, instruments, mood, and structure. Try three original examples and a simple method for revising your brief.
Read guide : How to write an AI music prompt: a practical brief and three examplesPut your next idea into music.
Bring a brief, a melody or a new direction to Mozart.
Create with MozartFrequently asked questions
Are stems the same as individual tracks?
The terms are sometimes used loosely, but a stem can combine multiple tracks into one audio file. A drum stem may contain kick, snare and cymbals together. If you need to change those instruments independently, ask for individual parts and confirm what each delivered file contains.
Can stem separation recover the original recordings exactly?
You should not assume that it can. Separation estimates components from a combined recording, while original session tracks existed before that mix was made. Listen for unwanted material or changes in the isolated part, especially in the section where you plan to feature it or raise its level.
Does stem separation create new music?
Its purpose is to isolate material from an existing recording. Generation creates new musical material. If you like a bass part but need access to it, separation may fit the task. If you want a different bass idea, look for a creation or replacement workflow instead.
Should I ask for dry or processed stems?
Ask the recipient what they need. Processed files preserve more of the intended sound; a dry alternative can provide flexibility for later decisions. Explain which effects are included and supply a reference mix when useful. Avoid assuming that an effect printed into audio remains independently adjustable.
Why should all stem files start at the same point?
A common start makes the files easier to align in another session, even when a particular part enters later. Agree on the render range and leave any necessary silence intact. After export, check that the files line up with a reference mix before treating the handoff as complete.
Are stems the same as MIDI?
No. Stems contain audio; MIDI describes performance information such as notes and their timing. A stem does not automatically retain editable notes or the instrument that produced them. Choose audio parts when you need recorded sound and investigate note-based workflows when you need to change the performance data.
How many stems should I request?
Request the number and grouping needed for the planned edits. A broad balance adjustment may need only a few groups, while a detailed mix may need individual instruments and vocals. Name the intended changes first, then agree on the file list rather than treating a higher count as inherently better.
What should I do if an isolated part sounds imperfect?
Compare it in the intended context as well as on its own. Decide whether the problem affects the actual edit, then consider a better source, different separation settings or access to original tracks. Keep the untouched source and document the limitation so another collaborator does not mistake it for an original recording.