Moises

What is Moises?

Moises is a music tool that takes a finished recording apart. Feed it a song and it returns the vocals, drums, bass and other instruments as separate tracks you can mute, solo or re-balance, along with the chords, the key and the tempo it detected along the way. The company describes it as the creative suite for musicians, built to help people practice, perform, create and collaborate from anywhere — and the ordering of those four verbs is a fair description of where the product's weight actually sits. Practice comes first.

That framing matters, because "vocal remover" is what most people search for and it undersells what is going on. Stripping the singer out of a track is one visible output of the separation model; the rest of the product is the scaffolding that makes an isolated track useful — slow it down without the pitch drifting, transpose it into a range you can actually sing, follow the chord chart as it scrolls, click along to a metronome that locked onto the song's own grid. A guitarist who wants a drumless version of a song to play over is using the same engine as a producer who wants a clean drum bus to resample.

The company behind it is Moises Systems, Inc., which trades as Music AI. Its patents page states plainly that every patent listed there is owned by "Moises Systems, Inc. d/b/a Music AI" — so Moises and Music.ai are two brands of one legal entity rather than a parent and a subsidiary, a distinction that is easy to get wrong from the website's footer alone. That same page provides something unusually checkable for a consumer AI product: virtual patent marking under 35 U.S.C. § 287(a), listing US 12,361,975, US 12,437,786, US 12,586,485 and US 12,586,606 as covering the consumer applications on iOS, Android, web, desktop and TV, and US 12,399,678 as covering the real-time Live product — numbers anyone can verify at the patent office rather than claims about "industry-leading technology." Music Business Worldwide reported that the company launched in 2019 under the Moises name and later split into a business-facing Music.ai platform and this consumer-facing Moises.ai product. Consumer scale is substantial: Apple's App Store shows a 4.70 rating across 27,486 ratings, and Google Play shows 4.5 stars across roughly 425,000 reviews in the 50M+ download bracket.

Core Features

• Stem separation into 2, 4 or 5 tracks: The stem separation engine is the technical foundation everything else rests on. The core engine splits an upload into your chosen number of stems and, in the same pass, detects the beat and the chords. Once the separation finishes you control each track's volume individually or mute it with one click, which is how both "karaoke version" and "drumless practice track" fall out of the same operation rather than being separate features.
• Advanced instrument separation on paid tiers: Beyond the basic vocals/drums/bass split, the store listings describe paid separation that pulls apart lead versus rhythm guitar, acoustic versus electric guitar, and main versus background vocals. This is the difference between "the guitars" as one blob and being able to isolate the part you are actually trying to learn.
• Chord detection with three difficulty levels: Chords are detected and displayed in sync with playback at easy, medium and advanced levels, with guitar chord shapes shown alongside lyrics where they occur. The graded levels are a teaching decision, not a technical one — a beginner gets a simplified chart of the same song an advanced player sees in full.
• Speed and pitch control with key detection: Tempo and pitch move independently, so a passage can be slowed to a workable speed without the pitch sagging, or transposed to fit a vocal range or an instrument's tuning. The tool detects BPM and key on its own and can transpose the detected chords into all twelve keys.
• AI Studio for arrangement: AI Studio generates new instrumental tracks from audio you upload, following your timing and matching your tone. It ships genre presets for drums, bass and guitar, expressiveness controls, and a Smart Selection Generator that lets you highlight any part of a track and have the system fill in the transition or groove. AI Mixing and AI Mastering finish the chain, with sample rate adjustable before export.
• Voice Studio for voice conversion: This is voice conversion rather than synthesis from scratch — the system keeps the melody, emotion and cadence of your original vocal performance and reframes it in a different singer's timbre, with pitch adjustable in semitones before you convert. The models are built from studio recordings by real, paid vocalists rather than scraped material.
• Lyric transcription and collaborative setlists: The app transcribes lyrics from audio in English, Spanish, Portuguese, French and Italian, and setlists let you invite collaborators and keep a band's material in one place. Note the transcription language list is finite — it is one of the few places where the product's coverage is narrower than its interface language count suggests.
• Moises Live for real-time control: A separate desktop-only application that runs entirely offline and applies separation to your computer's system audio in real time, so you can mute an instrument in a video that is already playing. It has no export function.

Use Cases

1. Learning a part by isolating it: The most common reason musicians open Moises. Mute everything except the part you are learning, slow it to a tempo you can actually play, loop the four bars that keep falling apart, and let the chord chart scroll underneath. The graded chord levels mean a student and their teacher can look at the same song at different levels of detail.
2. Building backing tracks for rehearsal or performance: Remove the instrument your band is missing — or remove yours to practice against the rest — and export the resulting mix. Because the tool accepts video files as well as audio, a live performance video from a streaming platform can become a rehearsal backing track without a separate extraction step.
3. Teaching music with recorded material: The company maintains a dedicated page for educators and says its tools are used by institutions including Berklee College of Music. In a classroom the practical value is being able to strip a recording down to whichever line the lesson is about, and to slow it without the recording turning into a chipmunk.
4. Sketching arrangements around an existing idea: AI Studio's premise is that you start with something you played, not with a text prompt. Upload a riff, generate a drum and bass part that follow your timing, regenerate the sections you do not like, keep the ones you do. It is arrangement assistance conditioned on your own audio rather than generation from nothing.
5. Reworking vocals without re-recording: Voice Studio converts a take you already sang into a different timbre while preserving your phrasing. For a songwriter demoing a track intended for a different type of voice, that removes the need to book a session singer at the demo stage.
6. Controlling audio you do not own the files for: Moises Live sits at the operating system level, so it can pull the dialogue up in a film or the vocals down in a music video that is streaming in a browser tab. Nothing gets exported; the point is listening control, not production.

How to use Moises

1. Create an account at moises.ai or in the mobile app. The free tier is available without payment, but note that the full pricing and feature table is only visible after you log in — the marketing site does not publish a public price list.
2. Upload a track. The supported input list is unusually explicit: audio in MP3, AC3, AAC, AIFC, OGG, WMA, AIFF, FLAC and WAV, and video in MP4, M4V, MOV, MKV, M4R, M4A, FLV, MPEG and WEBM. You can also import from a public URL or from cloud storage.
3. Choose your separation type — 2, 4 or 5 tracks. The system separates the audio and simultaneously detects the beat, chords and key, so by the time playback is ready you already have the analysis.
4. Work the mix. Mute or re-balance individual stems, switch the chord display between easy, medium and advanced, adjust tempo and pitch independently, set a count-in, and loop the section you are drilling.
5. Export what you need — individual stems or a combined mix, including the metronome click track if you want it. Desktop supports WAV export in addition to the compressed formats available elsewhere.
6. If you are creating rather than practicing, move into AI Studio to generate additional instrumental parts around your upload, then run AI Mixing and AI Mastering before your final export.

Tips & Best Practices

• Pick the separation count deliberately. A 2-stem split (vocals versus everything else) is cleaner than a 5-stem split of the same material, because every extra stem is another decision the model has to make. If you only need a karaoke track, do not ask for five stems and then wonder why the piano bleeds.
• Expect artifacts and plan around them. The company's own FAQ states that some level of quality change is expected with any vocal remover. Dense, heavily compressed mixes are the hardest cases. For practice this rarely matters; for a stem you intend to release, audition it in context before committing.
• Use pitch and tempo as separate controls. They are independent, so slowing a passage does not force the key down with it. When you are drilling something difficult, change one variable at a time so you know which one made the passage playable.
• Let the chord levels do the teaching. Starting a beginner on the advanced chord view is a fast way to lose them. Start at easy, get the changes under the fingers, then step up — the underlying detection is the same, only the display density changes.
• Give AI Studio a clear reference. It generates from your audio, so a clean, in-time riff produces better-fitting parts than a hesitant take with an ambiguous tempo. If you dislike a section, regenerate that section rather than the whole track.
• Understand where the paywall sits in Voice Studio. You can process an unlimited number of voice tracks without a subscription; what is limited is how many converted tracks you can export without Pro. Audition freely, then decide what is worth an export slot.
• Do not treat Moises Live as a substitute for the main app. The company itself says the original app delivers the highest-quality separation and recommends Live specifically for cases where real-time processing is required. Live also cannot export stems at all.
• Watch your credits if your plan uses them. The terms state that credits expire at the end of each monthly period, do not roll over, and are forfeited without compensation — including any left over if you downgrade or delete your account.

Who is Moises for?

• Instrumentalists learning repertoire: Guitarists, drummers, bassists and keyboard players who want to isolate or remove a part, slow it down and loop it. This is the product's centre of gravity and the reason its store ratings are as high as they are.
• Singers and vocal coaches: Key detection plus independent pitch shifting means a song can be moved into a comfortable range without the accompaniment being slowed. Removing the original vocal leaves an accompaniment to sing against.
• Music teachers and schools: The educator-facing material and the Google Play award in the personal-growth category both point at a real classroom user base. Graded chord levels and the shared setlist feature are the parts that matter most here.
• Producers and beatmakers: Advanced separation on paid tiers, Hi-Fi models and professional drum stem modules on Pro, plus VST plugin access, put the tool into a production workflow rather than only a practice one.
• Songwriters demoing ideas: AI Studio's conditioned generation and Voice Studio's timbre conversion both target the stage where an idea needs to be heard as a fuller arrangement before anyone commits studio time to it.
• Content creators working with existing music: Anyone who needs an instrumental bed, a cleaner dialogue track, or a specific instrument pulled out of source material they already have.
• Cover bands and worship teams: Setlists with collaborators, count-in control and the ability to build a track missing exactly the instrument the group lacks make this a rehearsal-room tool as much as a bedroom one.
• Listeners who want control, not production: Moises Live's audience is explicitly broader than musicians — the pitch is adjusting dialogue, effects and music levels in any app on your computer.

Platforms

• Web: Runs directly in the browser with no installation, and the company notes that certain advanced features are available on the web specifically.
• Desktop (macOS and Windows): A downloadable application. Desktop is also the only place WAV export is offered, per the store listings.
• iPhone and iPad: The iOS app requires iOS 17.0 or later, is a free download with in-app purchases, and carries a 12+ age rating. It was named Apple's iPad App of the Year for 2024 and was an Apple Design Awards finalist in 2025.
• Android: Listed on Google Play in the 50M+ download bracket with a rating of Everyone.
• Moises Live (desktop only): A separate Windows and Mac application, explicitly not available for mobile, running entirely on-device and offline.
• Localisation: The marketing site runs eighteen language editions, and the iOS listing declares 33 interface languages — worth separating from the lyric transcription feature, which covers only five languages.

Pricing & Plans

There is a free tier, and the company is direct about what it is: a limited number of uploads each month with access to a selected set of features. According to the store listings, that selection covers vocals, drums and bass separation, chord detection and the metronome click track, pitch shifting, BPM detection and setlists — enough to evaluate whether the separation quality suits your material. Full functionality and unlimited uploads require a paid subscription.

Above the free tier the structure is two paid levels. Premium adds unlimited separation and the advanced instrument splits (lead versus rhythm guitar, acoustic versus electric, main versus background vocals), fully unlocks the practice tools, and opens the advanced web features. Pro adds Hi-Fi separation models, professional stem modules including drum parts, uploads of up to 180 minutes of audio at a time, a fully unlocked Voice Studio and VST plugins. Note two contractual details before subscribing: only one free trial is permitted per user, and features bought à la carte are disabled and locked when the subscription period ends, which the terms acknowledge may leave you unable to reach some content you previously uploaded unless you resubscribe. This page deliberately quotes no prices — the site's own FAQ directs you to log in to see the full pricing and features table, so please check the current rates in the app rather than relying on figures republished elsewhere.

Alternatives

• LALAL.AI is the most direct comparison for separation quality alone, and sells credit packs rather than a practice-oriented subscription. If all you want is stems and you do not need chord charts or a metronome, the comparison is narrower and more favourable to it.
• Spleeter, Deezer's open-source separation library, is free and runs locally, which makes it attractive for batch work and for anyone unwilling to upload material to a cloud service. It provides no interface, no chord detection and no practice tooling — it is a component, not a product.
• Demucs, the open-source model family widely used in self-hosted setups, occupies similar territory: strong separation, zero convenience layer, and a hardware requirement you supply yourself.
• iZotope RX and similar repair suites approach the problem from audio restoration rather than practice, at a substantially different price point and with a far steeper learning curve.
• Standalone practice apps that slow down and loop audio without separating it remain a cheaper option if you only need tempo control and never need to remove a part.

The honest summary is that Moises competes on the bundle rather than on any single component. If you want only separation, cheaper and more specialised options exist; the case for Moises is that separation, chord detection, key and tempo control, transcription and mobile availability sit in one place.

Limitations & Considerations

• Separation is never lossless. The company's own FAQ concedes that some level of quality change is expected with any vocal remover. This is a physical limitation of the problem, not a defect specific to this product, but it means stems intended for release need auditioning rather than trusting.
• The public rating picture is genuinely split, and the split is informative. App Store shows 4.70 from 27,486 ratings and Google Play 4.5 from roughly 425,000 — both large enough to carry real weight. Trustpilot shows 1.5 from 54 reviews, with 78% one-star. Fifty-four reviews is a small sample, and Trustpilot itself flags two conditions on it: the profile is unclaimed, and the company has never invited customers to review, so the sample skews heavily toward people motivated to complain. Read the low score as a map of what goes wrong for unhappy users, not as a measurement of typical quality.
• The recurring complaints are commercial, not technical. The dominant Trustpilot themes are subscription entitlements changing — one reviewer reports features included in their original subscription being moved out of that tier mid-year with more money asked for the same functionality — and difficulty getting refunds or cancellations, including a refund refused within 24 hours and a user reporting months without a response on a cancellation. Similar "paid and it still asks for more" complaints appear in the Google Play reviews, where the company does reply publicly and points users to its help centre.
• No public price list. Pricing is only visible after login, which makes it harder to compare against alternatives before committing an account, and means any prices you find on third-party review sites may be stale.
• Credits do not roll over. The terms state credits expire monthly, do not carry forward, and are forfeited without compensation, including on downgrade or account deletion.
• Moises Live is deliberately limited. It cannot export stems, it is desktop-only, and the company states outright that the main app delivers higher separation quality.
• Lyric transcription covers five languages — English, Spanish, Portuguese, French and Italian — which is much narrower than the app's 33 interface languages.
• Training-data transparency has a stated boundary. The company markets itself on ethical, licensed training data, and its CEO told Music Business Worldwide that content providers include stock music services and studios working with indie artists across the US, Brazil and Europe — while acknowledging the company cannot disclose a full list of those providers. That is more disclosure than most competitors offer and still short of full auditability.
• Output rights come with restrictions. You own your uploads and your output, but the terms forbid redistributing or packaging that output as samples, sound effects, loops or part of a sample library even if modified, and forbid using it to compete with the service. Cover recordings remain your legal responsibility, and if you are signed, your label may have claims over recordings you make through the service.
• There is an age floor, and it is worth knowing before setting a class up. The terms require users to be at least 13 years old, and state that anyone aged 13 to 17 must have a parent or guardian who has affirmatively consented. The privacy policy matches this from the other direction: the company says it does not knowingly collect personal data from people under 13 or under whatever higher age limit applies locally, and asks anyone below that threshold not to use the service. For a school deployment this is the detail that determines which year groups can have their own accounts.
• Data handling is documented but not minimal. The privacy policy names concrete GDPR representatives — Instant EU GDPR Representative Ltd in Ireland for the EU and GDPRLocal Ltd for the UK — and describes pseudonymisation, encryption, access and retention policies, which is more specific than the usual boilerplate. Against that, Google Play's data safety section discloses that the app may share personal info and app activity plus two other categories with third parties, and may collect personal info, financial info and four further categories; data is encrypted in transit and deletion can be requested. Both things are true at once: the disclosures are unusually clear, and the collection surface is not small.
• No independent technical evaluation from mainstream tech press. Coverage exists, but it sits in music-industry trade publications and focuses on funding and business milestones rather than measured separation quality. There is no widely cited independent audio benchmark to point to.

FAQ

Q1. Is Moises free to use?

There is a genuine free tier, not just a trial. It provides a limited number of uploads each month and access to a selected set of features — the store listings describe that set as vocals, drums and bass separation, chord detection, the metronome click track, pitch shifting, BPM detection and setlists. Unlimited uploads and the full feature set require a paid subscription. The company also states that only one free trial is permitted per user, and warns that registering a new account to get around that limit may result in account termination.

Q2. How many stems can it separate a song into?

Two, four or five, chosen at upload time. Paid tiers extend this with advanced instrument separation that distinguishes lead from rhythm guitar, acoustic from electric guitar, and main from background vocals; the Pro tier adds professional stem modules including individual drum parts. As a rule of thumb, fewer stems means cleaner stems.

Q3. Can I use the output commercially?

You retain rights in your uploads and in the output you generate, and the company states its AI voice models are fully licensed and that finished tracks made with them are safe for commercial release. Three restrictions apply. You may not redistribute or package output as samples, sound effects, loops or part of a sample library, even modified. You may not resell or redistribute the voice models themselves or their isolated outputs. And separating a commercial recording does not give you rights to that recording — if you are working with someone else's song, the underlying licensing is your responsibility, and the terms say so explicitly for cover versions.

Q4. Does Moises train its AI on the audio I upload?

The company says no, in two places. Its site FAQ states plainly that Moises will never train on your data and that its models are trained only on licensed material with the rights holder's approval. The AI Studio page adds detail: audio you upload is used only for conditioning or inspiration during your session, then discarded. Its CEO has told the trade press that content providers include stock music services and studios working with independent artists in the US, Brazil and Europe, while noting the company cannot publish a full list.

Q5. What file formats and lengths can I upload?

Audio in MP3, AC3, AAC, AIFC, OGG, WMA, AIFF, FLAC and WAV; video in MP4, M4V, MOV, MKV, M4R, M4A, FLV, MPEG and WEBM. You can also import from a public URL or from cloud storage such as Google Drive, Dropbox or iCloud. On the Pro tier, single uploads can run up to 180 minutes of audio.

Q6. Which platforms does it run on?

Web in any browser, desktop applications for macOS and Windows, and native apps for iPhone, iPad and Android. Moises Live is a separate desktop-only application for Windows and Mac that runs offline; it is not available on mobile. The iOS app requires iOS 17.0 or later.

Q7. How good is the separation quality really?

Store ratings are strong at scale — 4.70 from 27,486 ratings on the App Store, 4.5 from roughly 425,000 on Google Play — and the company holds US patents covering the consumer applications and the real-time Live product. But no mainstream technology publication has published an independent measured evaluation, and the company itself concedes that some quality change is expected with any vocal remover. Practically: expect results that are more than good enough for practice and rehearsal, and audition carefully before using a stem in a release.

Q8. What do the negative reviews complain about?

Predominantly commercial issues rather than audio quality. On Trustpilot, where the score is 1.5 from 54 reviews with 78% at one star, the recurring themes are features moving out of a subscription tier mid-term with additional payment then required, refunds refused inside short windows, and difficulty cancelling. Two caveats on that sample: at 54 reviews it is small, and Trustpilot notes the profile is unclaimed and the company has never solicited reviews, so it over-represents users with grievances relative to the hundreds of thousands of store ratings.

Q9. Who actually makes Moises?

Moises Systems, Inc., trading as Music AI, headquartered at 136 S Main Street in Salt Lake City — an address that appears identically in the company's copyright-agent notice and in its Google Play developer information. The company launched in 2019 under the Moises name and now runs two brands: the business-facing Music.ai platform and this consumer product. In January 2025 it announced a $40 million Series A led by Connect Ventures, a CAA and NEA partnership, with Brazilian investor monashees, bringing total funding to $50.2 million.

Q10. How many people use it?

Three different numbers circulate and they measure different things, so it is worth keeping them apart. The company's own site says 70 million musicians. Music Business Worldwide reported 50 million registered users as of January 2025, up 66% from 30 million in April 2023. Google Play shows a 50M+ download bracket, which counts device installs rather than registered accounts. All three are consistent with a product at genuine consumer scale; none of them is independently audited.