Choose source audio and a mode for vocals, drums, bass, and other instrument stems.
4-Stem Split
Vocals, Drums, Bass, Other
Interface preview: separation, account library and example downloads are not connected. No file is uploaded.
A stem splitter estimates separate musical parts inside a finished mix, such as vocals, drums, bass, and the remaining instruments. These parts can support practice, arrangement study, or further editing. This workspace lets you prepare a source and mode; the actual separation service is not connected.
Choose the source, choose the output, and review the processing state.
Use a supported file you are allowed to process. Start with the cleanest available mix and confirm that it plays correctly.
Match the mode to your task. A simple vocal-and-backing split serves a different purpose from a multi-stem arrangement study.
The current interface confirms the disconnected backend when you submit. It does not simulate completed stems or provide a false download.
More control starts with a sensible source and realistic expectations.
A mix combines several parts into one recording. A multi-stem workflow attempts to recover useful layers so you can listen, mute, or work with them separately. The original recording still sets the ceiling for what can be recovered.
Use a two-part direction when you mainly need a voice or backing track. Choose a broader split when drums, bass, and instrumental layers matter individually. The most detailed option is not always the best choice for a simple practice task.
Listen to isolated parts and then to them together. Watch for softened transients, missing bass notes, vocal leakage, or unusual high-frequency texture. These checks help decide how much cleanup or rearrangement a future stem needs before it enters a larger production.
A clean mix is easier to work from than a clipped, noisy, heavily compressed, or repeatedly re-encoded file. Separation cannot reliably rebuild detail that is already gone. Keep a copy of the original and compare at matched listening levels.
Independent musical layers can be useful for listening, practice, and production planning.
Plan transitions or alternate arrangements around a vocal, drum part, or bassline. Check rights and sound quality before public use.
Study how the layers contribute to a groove, or assess a source for further processing. Separated audio remains an estimate of the mix.
Focus on the part you are learning or the accompaniment you need. Compare the result with the original to catch missing details.
Use a layer-by-layer view to discuss rhythm, arrangement, and phrasing. Treat it as an aid to listening, not a replacement for the original performance.
Modes, audio quality, downloads, rights, and the current service connection.
It estimates separate musical parts from a mixed recording. Unlike a multitrack session, those recovered stems were not recorded independently and may contain artifacts.
The interface prepares a source and a mode. Actual separation requires a backend provider, which is not connected in this preview; submitting displays that status.
The available modes describe the intended output groups. Choosing one prepares the request; it does not create stems while processing is disconnected.
A two-part vocal-and-accompaniment mode is the relevant workflow. Dense mixes and reverb can leave leakage, so any future result needs listening checks.
Follow the file formats and size or duration limits displayed in the workspace. A familiar extension does not guarantee that a damaged file can be decoded.
Quality depends on the recording and processing model. Expect possible leakage, missing detail, and changed texture, especially around reverb, distortion, and overlapping sounds.
No new stems are created in the disconnected preview. Download controls for real results can only become meaningful after a successful separation service is connected.
No. Processing a recording does not transfer rights to the composition, recording, or samples. Obtain the permissions required for the way you plan to use them.
The interface runs in your browser and lets you prepare local audio. The separation backend is currently disconnected, and the page explains that before any real processing job exists.