Suno Rebuilds Stem Separation From Scratch, Hitting 90% Accuracy
Suno's revamped stem separation now regenerates individual tracks from scratch, delivering artifact-free audio ready for professional DAW workflows.
- Regenerative approach: Suno now regenerates stems from scratch using the model's internal track representations, instead of frequency-based isolation.
- Three new modes: Auto Split (12 stems), Split from Mix (any single instrument), and Advanced Split (choose from ~100 instruments).
- Artifact-free output: Stems export as time-aligned 44.1kHz WAV files ready for Ableton, Logic, FL Studio, or any DAW.
- Tiered access: Pro subscribers get Auto Split and Split from Mix; Advanced Split is Premier-only. Credits range from 10–50 per extraction.
- Best for: Vocals and drums (90%+ accuracy); still struggles with dense orchestral layers, acoustic guitars, and backing vocals.
- Docs and demo: Full details at
Stem separation has always been the awkward middle step in AI music production: you generate something great, then watch it get mangled when you try to pull it apart. Suno just shipped a significant overhaul to its stem separation system that attacks this problem at the root, and the approach is meaningfully different from what came before.
From isolating to regenerating
Traditional stem separation works by analyzing a finished mix and trying to reverse-engineer the individual parts. Tools like Meta's Demucs or iZotope RX use spectral masks to predict which frequencies belong to which instrument. A modern stem separation tool converts audio into a spectrogram, predicts spectral masks for each sound source, and uses those masks to isolate vocals, drums, bass, and other instruments. The problem: when frequencies overlap between instruments, you get bleed, warbling, and phase artifacts that make the stems hard to use in a professional mix.
Suno's new approach sidesteps this entirely. Unlike basic stem separation tools that use frequency-based splitting, Suno's stem system has access to the generation model's internal representation of each instrument, which means cleaner separation with less crosstalk between stems. In other words, because Suno generated the track in the first place, it can regenerate each stem from scratch rather than trying to guess what was there. The result is stems that are cleaner, crisper, and free from artifacts, ready to drop into your track.
This is the key conceptual shift. The old paradigm was: mix first, separate later, and accept the losses. Until AI starts building tracks from the bottom up by creating stems first and mixing them into songs later, songs risk sounding synthetic. The fundamental problem is that AI needs to start at the start, rather than working backwards from the end. Suno's new system is a direct answer to that critique.
Three modes, three use cases
The updated Advanced Stem Separation breaks a song down with more precision and cleaner results than before, and it lets you decide how that happens with three unique separation modes: Auto Split, Split from Mix, and Advanced Split, each built for a different separation goal.
- Auto Split: The classic model, splitting songs into 12 stem categories. Great for quick separation. Costs 50 credits per extraction.
- Split from Mix: Updated from the old Vocal + Instrumental mode, now with the ability to choose which instrument you split, not just vocal. Costs 10 credits per extraction (20 total for both resulting stems).
- Advanced Split: Lets you decide which stems your track gets separated into. Choose from a list of nearly 100 instruments and create custom stems to suit your track. Also 10 credits per stem extraction. Available for Premier subscribers only.
What it's good at, and where it still struggles
Stem quality varies by instrument type. Drums and bass tend to separate very cleanly because they occupy distinct frequency ranges. Vocals separate well because the AI model has a clear internal representation of the vocal track. In hands-on testing, separation accuracy for vocals and drums exceeded 90%. However, complex orchestral arrangements and densely layered instrumental passages still have room for improvement.
Where it can fall down in terms of quality are things like acoustic guitars, piano tracks, string lines, and backing vocals. These are usually the areas that require the most attention. A useful heuristic: the more harmonically distinct an instrument is from the rest of the mix, the cleaner the stem will be.
There's also a practical gotcha worth knowing: stem separation is a paid feature. Pro subscribers get Auto Split and Split from Mix. Premier subscribers get those two options, plus Advanced mode.
The DAW workflow this unlocks
The practical upside for producers is real. You can export up to 12 time-aligned WAV stems and use them seamlessly in Ableton, Logic, or any DAW. Clean, structured, and ready for pro workflows. The stems export at 44.1kHz, and the files are properly time-aligned, meaning they line up perfectly when imported into a DAW at the correct BPM. Each stem starts at the same point in time as the original track, so there is no manual alignment needed.
The most common production pattern emerging from the community: use Suno to generate the compositional skeleton, pull the stems, then replace or augment individual elements with live recordings or your own samples. Many producers use Suno for initial song ideas, then replace the drums entirely with their own samples and re-record bass, keeping the AI vocals and melodic elements as the creative core.
To access the feature, select Get Stems from the Edit menu and start extracting. The full breakdown of modes and credit costs is covered in
Comments
