Cubase 15 & Nuendo 15 Stem Separation in 2026: AI Stem Splitter, Native Cubase Stem Separation & SpectraLayers Compared

Cubase 15 (November 2025) added native AI Stem Separation to the DAW — vocals, drums, bass, other, four stems, free with Cubase Pro. Nuendo 15 (March 2026) inherits the same engine plus deeper post-production tooling. Both ship with SpectraLayers Go bundled, with an optional upgrade to SpectraLayers Pro 12 ($349.99) that unlocks Unmix Song, Unmix Drums, and the post-production-focused Unmix Soundtrack.

There are three practical answers in 2026 — plus iZotope RX as a known third-party companion (with one important caveat for Steinberg users).

The short answer:

• For the cleanest possible stems on modern music → external AI (StemSplit) at $0.10/min — drag the WAVs straight into your Cubase/Nuendo project.
• For the fastest in-DAW workflow → Cubase 15 native AI Stem Separation (Cubase Pro 15 / Nuendo 15). Free with the DAW. Vocals, drums, bass, other.
• For deep spectral work → SpectraLayers — Go bundled, Pro upgrade ($349.99) for Unmix Song, Unmix Drums, and the killer post feature Unmix Soundtrack (dialogue / music / effects from a single mix).
• For pro post production → Nuendo 15 + SpectraLayers Pro 12. Unmix Soundtrack does what iZotope Scene Rebalance does, with deeper ARA 2 integration than Pro Tools can match.
• For external repair → iZotope RX 12 via VST3 inserts or RX Connect. Important caveat: RX 12 Spectral Editor ARA is not supported in Cubase/Nuendo in 2026 (only Logic Pro and Studio One are on iZotope's ARA host list). Use SpectraLayers for in-DAW spectral instead.

Cubase has supported ARA 2 for years — and SpectraLayers' tight ARA integration with Cubase/Nuendo is one of the best implementations of any spectral tool in any DAW. This guide walks through every method honestly and tells you which to use when.

Try StemSplit on your Cubase project →

Method 1: StemSplit (External AI, Best Quality, Drag-and-Drop into Project)

This is the cleanest, fastest, cheapest option. Works on every Cubase and Nuendo edition — Elements, Artist, Pro, Nuendo full. Browser-based; you bring the separated WAVs back to the project as audio events on new tracks.

Workflow

1. Bounce the source from Cubase — File → Export → Audio Mixdown → WAV at project sample rate. Or upload the original file directly without bouncing.
2. Upload to StemSplit → choose 4-stem (vocals, drums, bass, other) or 2-stem (vocals + instrumental).
3. Download the stems as WAV files.
4. Drag the WAVs into Cubase — drop directly onto the Project window at the original timecode. Group them into a Folder Track for clean organisation.

The model is htdemucs FT — ~8.4 dB SDR in published benchmarks. Cleaner than Cubase 15's native Stem Separation on dense modern productions, and competitive with SpectraLayers Pro's Unmix Song on most material.

When to use this method

• You need release-grade stems for music mixes, sample packs, references, or remix work
• You're on Cubase Elements or Artist (no native Stem Separation in those tiers)
• You want the absolute peak quality, faster than running the engine on your CPU
• One-off jobs where you don't want to commit a SpectraLayers Pro upgrade
• iPad/laptop sessions where SpectraLayers isn't available

Pricing reality check

$0.10/minute. A 3-minute song = $0.30 in credits. Free 30-second preview before you pay. About 3,500× cheaper than SpectraLayers Pro 12 ($349.99) per song, and the quality is at the top of what's publicly available.

Cubase / Nuendo-specific tips

• Match sample rate to the project before separating to avoid resampling artifacts (most music projects are 44.1 kHz; post is typically 48 kHz; high-end music sometimes 96 kHz).
• Use Folder Tracks — drop the four returned stems into a Folder Track to keep mute/solo/level grouped. Cubase's nested folder behaviour makes complex stem mixing tractable.
• Bus the stems through a single Group Channel for parallel processing or to print master-bus FX onto the stems.
• Use Render In Place to commit any post-import processing and keep the stem set portable.
• Snap to original timecode when importing — File → Import → Audio File → preserve embedded timecode if you bounced as BWAV.

---

Method 2: Native AI Stem Separation in Cubase 15 / Nuendo 15

Steinberg added native AI Stem Separation in Cubase 15 (November 2025); Nuendo 15 (March 2026) inherits the same feature. Free with the DAW at Pro tier and above. Four stems: Vocals, Drums, Bass, Other.

Workflow

1. Drop the source audio onto a track in your project
2. Right-click the audio event → Stem Separation (or Audio menu → Stem Separation in Cubase Pro 15)
3. Cubase processes the file locally on your CPU and creates four new tracks — Vocals, Drums, Bass, Other — with a Folder Track grouping them
4. The original event is muted; you can solo, mix, and process the stems individually

Edition gating (2026)

If you're on Elements or Artist, you'll need StemSplit (Method 1), SpectraLayers Pro standalone (Method 3), or a Cubase Pro upgrade.

When the native feature is the right answer

• You're on Cubase Pro 15 or Nuendo 15 and want zero setup
• You don't need release-grade quality — drafts, references, samples for sketching
• You're working offline / privacy-sensitive material
• You want stems automatically grouped on the timeline at the original position

Honest limitations vs StemSplit

• Quality is solid but not class-leading — htdemucs FT (StemSplit) outperforms the native Cubase engine on dense modern productions
• Locked to four stems (vocals, drums, bass, other) — no piano/guitar/2-stem modes
• Processing time scales with track length and CPU — long files can be slow on older machines
• Requires you to bounce instruments and virtual sources to audio first

For drafts and sketches, native is genuinely fast. For final-grade stems on modern music, StemSplit wins.

Native Stem Separation vs SpectraLayers — what's the relationship?

Steinberg's official answer (and the most discussed forum question after Cubase 15's launch): they coexist. The native Cubase Stem Separation is a fast, simple, four-stem unmix tied to the timeline event. SpectraLayers Pro's Unmix Song offers more stems (up to six: vocals, drums, bass, guitar, piano, other), more control, and additional Unmix modes (Drums, Soundtrack, Instrument). Use native for speed; switch to SpectraLayers when you need more granularity or post-production-specific separation.

---

Method 3: SpectraLayers Go (Bundled) and SpectraLayers Pro 12 (Upgrade)

SpectraLayers is Steinberg's spectral editor with built-in AI source separation. Cubase 15 and Nuendo 15 include SpectraLayers Go out of the box; SpectraLayers Pro 12 ($349.99 standalone) is the upgrade unlock for the full feature set.

What SpectraLayers Pro 12 offers

ARA 2 integration with Cubase / Nuendo

SpectraLayers' deepest advantage is its ARA 2 integration with Cubase and Nuendo — one of the best in any DAW. Right-click an audio event → Extensions → SpectraLayers (ARA). The event opens in SpectraLayers as a non-destructive layer on the timeline. You can:

• Edit horizontally across multiple clips in ARA mode
• Group by track for batch operations
• Audition unmix layers individually before committing
• Re-open the same event later to revisit edits — they persist with the project

This makes SpectraLayers the de facto in-DAW spectral / unmix tool for Cubase and Nuendo workflows. It's also why iZotope RX's lack of ARA Spectral Editor support for Steinberg DAWs (Method 5) matters less than it does on a Pro Tools-style workflow.

Pricing

If you're on Cubase Pro 15 or Nuendo 15, you already have SpectraLayers Go. Most pro post houses and music engineers running deep spectral work upgrade to Pro 12.

When SpectraLayers is the right answer

• You want six-stem unmix (vocals, drums, bass, guitar, piano, other) instead of four
• You need Unmix Drums to split a drum bus into kick/snare/hats individually
• You're doing post production and need Unmix Soundtrack (dialogue/music/effects) — see Method 4
• You want non-destructive, ARA-based spectral editing on the Cubase/Nuendo timeline
• You're cleaning up dialogue / voice content (Unmix Noisy Speech)

Honest limitations vs StemSplit

• For pure stem separation quality on modern music, StemSplit (htdemucs FT) is consistently as clean or cleaner than SpectraLayers Pro's Unmix Song
• The $349.99 entry for Pro 12 only makes sense if you also want the other modules (Drums, Soundtrack, etc.)
• Processing happens on your CPU — a 6-minute song with Unmix Song at Pro quality can take meaningful time

For Cubase/Nuendo users who already have Pro 12, SpectraLayers + StemSplit is the strongest pairing: SpectraLayers for non-destructive spectral work and post-specific unmixing, StemSplit for the cleanest possible music stems.

---

Method 4: Nuendo 15 + SpectraLayers Pro 12 — The Pro Tools Alternative for Post

This is the section for post-production engineers. Nuendo 15 (March 2026) is Steinberg's flagship for film, TV, broadcast, and game audio post — the direct competitor to Pro Tools. Combined with SpectraLayers Pro 12, it offers a workflow Pro Tools engineers can't match natively.

What Nuendo 15 brings to post

• Analyzer Track — dialogue intelligibility metering (Fraunhofer IDMT) plus LUFS
• Convert Channels — fast surround/immersive routing
• MXF OP1a import, video export improvements
• Folder Tracks with integrated group-channel behaviour
• ADR panel refinements (deep native ADR is Nuendo's traditional strength vs Pro Tools)
• Headphones Match curve updates
• WaveLab Go + SpectraLayers Go bundled
• Native AI Stem Separation (same engine as Cubase 15)

Unmix Soundtrack vs iZotope Scene Rebalance

If you've been a Pro Tools post engineer evaluating iZotope RX 12 Advanced for Scene Rebalance ($1,399), Steinberg's answer in Nuendo 15 + SpectraLayers Pro 12 is Unmix Soundtrack:

Pro Tools wins on real-time AAX adjustment during a mix review (RX 12's April 2026 update). Nuendo wins on integrated DAW + spectral tool cost, ARA 2 depth, and native ADR / video / immersive workflow. For broadcast and game audio in particular, Nuendo + SpectraLayers Pro is increasingly the answer over the Pro Tools + RX combo.

Workflow: Unmix Soundtrack in Nuendo

1. Drop the mixed scene clip onto a track in Nuendo 15
2. Right-click the event → Extensions → SpectraLayers (ARA)
3. In SpectraLayers Pro 12 → Unmix Soundtrack module
4. SpectraLayers separates dialogue, music, and effects layers
5. Adjust each layer's level, mute, or solo — non-destructive in ARA
6. Close SpectraLayers — edits persist on the event in Nuendo

For dialogue editing specifically, Unmix Noisy Speech in the same SpectraLayers session lets you isolate breath, mouth noise, and broadband background from voice — workflows that would otherwise need RX Voice De-noise + Mouth De-click.

---

Method 5: iZotope RX 12 in Cubase / Nuendo (with One Important Caveat)

iZotope RX is widely used in Cubase and Nuendo workflows — but unlike Logic Pro and Studio One, Cubase and Nuendo are not on iZotope's RX 12 ARA host list for the Spectral Editor. In 2026 the RX 12 spec sheet only shows AU ARA Spectral Editor for Logic 10.7+ and VST3 ARA for Studio One 7 / Fender Studio Pro 8.

What this means in practice

• RX 12 Music Rebalance: run as a VST3 plug-in insert on the track. Adjust the four stem sliders and render via Render In Place / Audio Mixdown. New in RX 12: real-time AAX exists, but it's Pro Tools-only — not VST3.
• RX 12 Scene Rebalance: same — VST3 insert or standalone via RX Connect.
• RX 12 Spectral Editor: use standalone RX via RX Connect. Track → Render or Bounce, send the audio to RX standalone, edit, return to Cubase. Not as fluid as the ARA workflow Logic / Studio One get.
• RX repair plug-ins (de-click, de-noise, de-ess): all available as VST3 inserts in Cubase / Nuendo. Standard plugin workflow.

Pricing recap

Should Cubase / Nuendo users buy iZotope RX in 2026?

• For repair plug-ins (de-click, de-noise, de-ess, mouth de-click, breath control): yes — they work as standard VST3 inserts and are the industry standard
• For Spectral Editor: maybe not — SpectraLayers Pro 12 (already bundled as Go, $349.99 upgrade) does the same job with better ARA 2 integration in Cubase/Nuendo
• For Music Rebalance: probably not — StemSplit is cleaner and 1,300× cheaper per song
• For Scene Rebalance specifically: SpectraLayers Pro 12's Unmix Soundtrack is the cheaper, ARA-integrated alternative

The recommended Cubase/Nuendo + RX combo in 2026 is: RX 12 Standard for the repair plug-ins, plus SpectraLayers Pro 12 for spectral and unmix work, plus StemSplit for peak-quality music separation.

---

Method 6: Manual Workarounds (Free, Limited)

If you don't have any of the above and need to extract or remove vocals on a budget, Cubase's built-in plug-ins can do limited "vocal reduction." Honest assessment: not competitive with AI on modern music.

Phase cancellation

1. Duplicate the stereo audio onto a new track
2. Convert to dual mono (Audio → Convert) and invert one channel (or use UR / Stereo Enhancer to invert)
3. Sum to mono — centred content cancels (vocal, kick, bass)

Why it fails: vocals are wide, so they don't fully cancel. Bass and kick are centred and cancel along with the vocal. Result: leaky, bass-depleted mix.

Spectral notch with StudioEQ / Frequency 2

1. Insert Frequency 2 on the source
2. Cut a wide notch at 1–4 kHz (vocal presence) by -6 to -12 dB
3. Vocal becomes quieter; rest of the mix gets hollow

For real spectral work, use SpectraLayers (Method 3). Stock EQ alone won't get clean results.

---

Method Comparison

For most Cubase music engineers, the right pairing is:

• Cubase Pro 15 native Stem Separation for fast in-session work
• StemSplit for peak quality on release stems
• SpectraLayers Pro 12 if you regularly need 6-stem or surgical spectral

For Nuendo post engineers:

• Nuendo 15 + SpectraLayers Pro 12 as the primary post stack
• Unmix Soundtrack for dialogue/music/effects rebalancing
• StemSplit for peak music stems and quick external prep

---

Cubase / Nuendo-Specific Tips

Use Folder Tracks for stems. When you import four stems separated externally (or eight via Unmix Drums), drop them into a Folder Track for grouped mute/solo and a single composite fader. Cubase's nested folder behaviour scales well — Stems folder → Drums sub-folder → individual elements.

Match sample rate before separating. Most music projects are 44.1 kHz; Nuendo post is typically 48 kHz; some music is 96 kHz. Mismatched rates introduce resampling artifacts. Bounce at project rate, separate, import — no rate conversion in between.

Use Render In Place for commit. When you've rebalanced a track via SpectraLayers ARA or Music Rebalance VST3, Render In Place commits the result to a new audio event, freeing CPU and making the session portable.

ARA 2 + SpectraLayers is your friend. Right-click → Extensions → SpectraLayers is faster than RX Connect for any spectral or unmix work in Cubase/Nuendo. Treat SpectraLayers as the in-DAW spectral default; reach for RX standalone only for specific repair tasks.

Bounce flat before separating externally. Disable Insert FX, clip gain, and channel processing on the source bounce before sending to StemSplit — pre-bounce processing leaks into the AI's view of the mix.

For Nuendo: use ADR Marker tracks when prepping stems for a dialogue replacement workflow. Markers from the original ADR session can guide where you need to apply stem-specific repair.

---

Frequently Asked Questions

Does Cubase 15 have built-in AI stem separation?

Yes — Cubase 15 (November 2025) added native AI Stem Separation to Cubase Pro. Right-click an audio event → Stem Separation. Produces four stems: vocals, drums, bass, other. Cubase Elements 15 ($99.99) and Cubase Artist 15 ($329) do not include the feature; Cubase Pro 15 ($579) and Nuendo 15 (~$999.99) do.

What's the difference between Cubase Stem Separation and SpectraLayers Unmix Song?

The native Cubase Stem Separation is a fast, simple, four-stem unmix (vocals, drums, bass, other) tied to a timeline event. SpectraLayers Pro 12 Unmix Song offers up to six stems (vocals, drums, bass, guitar, piano, other) plus more control and additional unmix modes (Drums, Soundtrack, Instrument, Noisy Speech). Use the native feature for speed; switch to SpectraLayers when you need more granularity or post-production-specific separation.

What's the best stem splitter for Cubase in 2026?

For peak quality on modern music, StemSplit produces the cleanest stems — htdemucs FT consistently outperforms the native Cubase 15 engine and matches or beats SpectraLayers Pro on dense modern productions, at $0.10/min. For zero-setup in-DAW work, Cubase 15 native Stem Separation is the best free option (Pro tier and above). For 6-stem unmixing or post-production unmix tasks, SpectraLayers Pro 12 ($349.99 upgrade from the bundled Go).

Is SpectraLayers Pro 12 included with Cubase Pro 15?

No — Cubase Pro 15 and Nuendo 15 include SpectraLayers Go, the streamlined bundled tier. SpectraLayers Pro 12 (the full version with Unmix Song, Unmix Drums, Unmix Soundtrack, etc.) is a separate $349.99 upgrade. If you only need basic spectral editing and ARA 2 integration, Go is enough; for stem separation specifically, you'll want Pro.

Does SpectraLayers integrate with Cubase via ARA 2?

Yes — SpectraLayers' ARA 2 integration with Cubase and Nuendo is one of the deepest of any spectral tool in any DAW. Right-click an audio event → Extensions → SpectraLayers. Edits are non-destructive, persist with the project, and can extend horizontally across multiple clips.

Does iZotope RX 12 Spectral Editor work as ARA in Cubase or Nuendo?

No — iZotope's published RX 12 ARA host list in 2026 only includes Logic Pro 10.7+ (AU ARA) and Studio One 7 / Fender Studio Pro 8 (VST3 ARA). For Cubase and Nuendo users, RX 12 Spectral Editor runs only as a standalone application via RX Connect, or RX repair plug-ins as standard VST3 inserts. Use SpectraLayers Pro 12 instead for ARA-based spectral work in Cubase/Nuendo.

How much does Cubase 15 cost?

Cubase Elements 15 ~$99.99, Cubase Artist 15 ~$329, Cubase Pro 15 ~$579. Crossgrade, EDU, and Yamaha bundle pricing can change net price meaningfully. Steinberg occasionally runs version-launch promotions; check official pages and Sweetwater/Thomann at purchase time.

How much does Nuendo 15 cost?

Nuendo 15 lists at approximately $999.99 USD for the full version. Crossgrade pricing from older Nuendo versions or from Cubase Pro is available — check Steinberg's upgrade pages.

Is Nuendo or Pro Tools better for post production in 2026?

It depends on your facility and pipeline. Pro Tools remains dominant in US film theatrical mixing and session interchange; Nuendo is widely used in European broadcast, game audio, and immersive post, and increasingly in TV post. The 2026 inflection point is that Nuendo 15 + SpectraLayers Pro 12 offers Unmix Soundtrack (dialogue/music/effects) for $1,349.99 perpetual vs Pro Tools + RX 12 Advanced at ~$1,698/year subscription. For new post setups not bound to Pro Tools session interchange, Nuendo's value is increasingly competitive.

Will Cubase / Nuendo add real-time AI stem separation?

The native Cubase 15 / Nuendo 15 Stem Separation is offline by design — it processes a source clip and creates new tracks. iZotope RX 12 added real-time AAX Music Rebalance for Pro Tools in April 2026, but no equivalent VST3 real-time stem rebalance has been announced for Cubase/Nuendo as of May 2026. SpectraLayers' ARA-based unmix is the closest non-destructive equivalent.

What sample rate should I bounce from Cubase for stem separation?

Match your project rate. 44.1 kHz is the music standard; 48 kHz is standard for film/TV/broadcast post (Nuendo); 96 kHz for high-end music sessions. Don't upsample before separation — it doesn't add information. Both StemSplit, native Cubase Stem Separation, and SpectraLayers Pro handle 44.1 / 48 / 88.2 / 96 kHz correctly.

---

Get Production-Grade Stems for Your Cubase or Nuendo Project

Upload any track to StemSplit and drag clean WAV stems straight into your Cubase or Nuendo project — Elements, Artist, Pro, or Nuendo full.

• Vocals, drums, bass, and other — as separate WAV files
• Cleaner than Cubase 15 native Stem Separation on dense modern productions
• Works on every Cubase and Nuendo edition (and every legacy version)
• $0.10/min — about $0.30 for a typical song
• Free 30-second preview before you pay
• No plug-in install, no upgrade required

Separate Stems Now →