Mubert API vs SOUNDRAW API: Which Workflow Fits In-App Soundtracks?
Last verified: October 19, 2026. Every price, allowance and licence clause below was read that day on the vendor’s own site; every source link at the end resolves. Two gaps are declared rather than papered over: Mubert’s master API contract is not published (only plan-level feature labels are), and neither API was put through a load test, so Score withheld. Where a marketing page and a legal document disagree, both are quoted and the disagreement is left standing.
A product team comparing Mubert and SOUNDRAW usually starts with the demo track. That is the wrong artefact. The thing that decides whether your soundtrack feature can ship is not which generator returns the nicer loop — it is whether the music reaches your user as playback, as a downloaded file, or wrapped inside a video they export. Those three shapes sit under three different grants, and the two vendors draw the boundaries in almost opposite places.
This comparison is written for the person who has to defend the integration later: what the API asks you to build, what the contract lets your users do with the output, and what changes when a prototype becomes a product. If you are still choosing between four vendors rather than two, the wider survey in our AI music API comparison covers the field first; this piece assumes Mubert and SOUNDRAW are the two finalists.
Three delivery shapes, three different clauses
Write down which of these you are building before you compare anything else, because the answers stop being interchangeable at exactly this point.
- Playback inside the app. The user hears music; no file leaves your system. This is closest to “distributing an application”, and both vendors have language for it — but only one sells the sub-licensing that user-facing playback implies.
- Export of the audio file. The user downloads a track and keeps it. This is distribution of the asset itself, and it is the shape both contracts treat most cautiously — Mubert’s published plans exclude distribution even at the top tier, and SOUNDRAW routes it into a revenue-share licence.
- Music bed inside a user-made video. The user exports a video, not audio. This is the shape most creator tools actually ship, and it is the one place where SOUNDRAW’s public terms are unusually explicit about what counts as enough transformation.
How the two APIs are shaped
Mubert: two credential tiers and a licence object you can query
Mubert’s v3 API is JSON over HTTPS, split into two directories that behave like two different products. /service/ is yours: it runs on a company-id plus license-token pair and manages licences and customers. /public/ is what your end users’ sessions run on: a customer-id plus access-token pair, used for generation, library search and streaming. Streaming URLs live on a separate host from the REST calls.
The licence is a readable object rather than a PDF promise. A GET /service/licenses call returns the features the plan actually carries (track, ttm, itm, streaming), the permitted bitrates, the permitted output formats (mp3 and wav in the documented example), the permitted modes (track, jingle, loop, mix), a maximum track duration, and counters for monthly and daily generation plus streaming minutes, alongside a customers_count and its limit. Limits set to -1 mean uncapped. That structure matters for engineering: quota, seat count and output shape are configuration your code can read at runtime instead of assumptions baked into a release.
Per-user separation is a first-class concept. Each new customer you provision consumes one seat against customers_count_limit; the docs say a fresh client key “creates a customer and uses one of your seats”, so a naive retry that generates a new key per session silently burns your seat allowance. Generation status can be pushed rather than polled: a PUT on the licence sets a webhook_url, after which Mubert posts track models as they finish. For latency, the track store pre-renders buckets at fixed durations (5, 6, 8, 10, 15, 20, 30, 40, 60, 180, 240 and 300 seconds in the documented list) and returns those in a single request, while any other length is rendered from scratch — but the store has to be enabled under your contract first, which means an email to their team rather than a dashboard toggle.
Two operational details are easy to miss and expensive to discover in production. Track URLs expire: the sample licence object in the docs carries track_expiration_time: 900, and the docs tell you to save the file you intend to keep. And a stream URL is itself a credential — one stream per client, with a second stream replacing the first, so two players in one session will fight each other.
SOUNDRAW: one token, and a legal document that does the heavy lifting
SOUNDRAW’s API page sells the capability set rather than the plumbing: generation driven by genre, mood and theme inputs, embedding inside apps, games, video platforms and multi-user products, and a direct answer to the obvious question — “Can multiple users generate music through my app? Yes. The API is built for multi-user environments and scalable generation.” What the public page does not give you is an endpoint reference, a rate-limit table or a status page; the technical documentation is handed over after you become a customer, and the API documentation link on the page resolves to the commercial agreement, not to a developer portal.
Its published entry points are unusually concrete for a sales-led API. API Starter is $29.99 per month for up to 100 songs, described as being for indie developers and companies of up to three employees and marked exclusive to new signups. API Pro is $300 per month for up to 1,000 songs with a six-month minimum commitment. The same page’s FAQ then says pricing “is typically based on usage … for exact pricing, contact the SOUNDRAW team”, which contradicts the two prices printed directly above it — treat the numbers as entry points to be confirmed at checkout, not as a rate card.
The contract is where SOUNDRAW is more explicit than Mubert, and it is also where the integration risk concentrates. The API General Agreement grants a “non-exclusive, non-transferable, non-sublicensable, worldwide, revocable” right to call the API and to “develop, implement and distribute Applications solely for use by its users and subscribers”. Section 3.5 then closes the obvious loophole: “Sublicensing of the assets to third parties is prohibited.” Section 3.10 makes the token non-shareable, and section 3.14 defers rate limits to whatever SOUNDRAW notifies you in writing — there is no published ceiling to design against.
Two clauses change architecture more than any feature. Download links expire 72 hours after generation (section 3.17), so any “save to my library” feature must copy the file into your own storage inside that window. And section 3.16 forbids caching service data beyond a reasonable time and requires refreshing data obtained via the API at least once per day — a clause written for data APIs that lands oddly on audio, and one worth clarifying in writing before you build a permanent asset store on top of it.
Where the licence stops
This is the part that separates the two vendors, and it is not visible from either pricing page at a glance.
Mubert sells sub-licensing as a plan feature — at the third tier
The API plans page lists the same capability grid for all three self-serve tiers, and the differences are almost entirely legal rather than technical. Trial ($99 list, $49 in the sale running on October 19, 2026) gives 100 generations and 100 streaming minutes with text-to-music and image-to-music, and explicitly withholds monetized content, the curated library, sub-licensing and distribution. Startup ($249 list, $199 in the sale) adds 5,000 generations, 5,000 streaming minutes, the 12,000-track library, webhooks, lossless output and monetized content — and still withholds sub-licensing and distribution. Only Startup+ ($999 list, $499 in the sale, 30,000 generations and 30,000 streaming minutes) turns sub-licensing on, and even there “No distribution” remains listed. A custom plan covers vocals, stems, audio branding, music schedules and advertising.
Read that as an integration constraint: the tier that lets your users hear generated music as part of your product is the $499–$999 tier, not the $199 one. A team that prototypes on Startup and plans to “flip on sub-licensing later” is planning a plan change, not a config change. The same grid says no published plan includes distribution, so an export-to-file feature is a negotiated custom contract on Mubert’s side regardless of volume. Sale prices carry a countdown on the page and are non-refundable; the list prices above are the figures that will outlive the promotion.
SOUNDRAW prohibits sub-licensing and prices the exception as revenue share
SOUNDRAW’s route is different in kind. The consumer licence page grants personal and commercial use across all plans — YouTube, UGC, product videos, tutorials, live streams, social, video games, up to one physical store, client work, hold music, radio and TV — and lists what is not permitted: content not owned by the user (movies, TV, ads), theatre and other synchronised content, Content ID registration, distribution without modification, and reselling a modified beat on stock platforms. That page is written for Creator and Artist subscribers, and the Terms of Service sharpen it: the licence is “non-exclusive, non-transferable, and non-sublicensable”, and Article 10 forbids “selling, distributing without charge, or granting a sublicense for the Generated Music (excluding videos and other deliverables created using said Generated Music) without the Company’s prior consent”.
The API agreement replaces that prohibition with a priced exception. If you let end users download SOUNDRAW-generated songs without substantially transforming them, the Extended License applies: 70% of the revenue from those sales goes to SOUNDRAW, net of platform fees; if the songs are significantly altered, 50%. Payment is quarterly, with accounting statements, audit rights on 60 days’ notice and an accredited auditor in Japan. Section 4.4 also sets a transformation floor for user content — a new sound covering 60% or more of the track, and the result must clearly sound like a different song — while carving out videos whose main purpose is not the music.
So the two vendors solve the same problem in opposite directions. Mubert pre-prices the right to sublicense and puts it in a tier. SOUNDRAW withholds sublicensing by default and turns it into a negotiated revenue share. For a subscription product where the music is incidental, SOUNDRAW’s structure is cheaper and clearer; for a product whose whole value is handing users music, Mubert’s tier is the more predictable cost.
The decision table
| Decision | Mubert API | SOUNDRAW API |
|---|---|---|
| Ship a working prototype this month | Trial at $49–$99/month: 100 generations, no monetized content, no library — enough to prove the flow, not to launch | Starter at $29.99/month: 100 songs, capped at companies of three employees, new signups only |
| Launch user-facing generation at roughly 5,000 tracks a month | Startup at $199–$249: quota and library fit, but sub-licensing is still off | Pro at $300/month for 1,000 songs, six-month minimum — 5,000 songs is outside the published table |
| Let users hear music in-app without downloading | Startup+ at $499–$999 is the first tier that grants sub-licensing | Covered by the application grant in section 2.2, but 3.5 bars sublicensing — get the delivery model named in the signed agreement |
| Let users export the audio file | No published plan permits distribution, including Startup+ — custom contract required | Extended License: 70% revenue share unaltered, 50% significantly altered, quarterly reporting and audit |
| Music inside a video the user exports | Monetized content from Startup upward; sub-licensing from Startup+ | Section 4.4 allows it when music is not the main purpose and the 60% transformation test is met |
| Per-end-user quotas, seats and generation callbacks | Native: customer credentials, customers_count limit, webhooks, per-licence counters |
No published per-user provisioning model or webhook specification — treat as unverified until the technical docs arrive |
| Exit the contract and keep shipping what you built | Contract text not public: plan labels only, and “No distribution” is the one hard signal | Section 13.2: 30 days’ notice either way, and content already integrated stays licensed perpetually; 13.6 keeps saved tracks usable |
From prototype to production: where the handoff breaks
Three failures recur when teams move from a demo to a shipped feature, and both vendors have a version of each. The same handoff problems surface in audio-only comparisons too: our Eleven Music and Stable Audio API review ran into the same expiry and export questions with a different pair of contracts.
The tier you demoed on cannot legally ship. On Mubert this is structural: Trial and Startup both withhold sub-licensing, so the moment another human being hears generated music inside your product, you are on Startup+ or a custom deal. On SOUNDRAW the equivalent trap is the employee cap — Starter is scoped to companies of up to three people, and an organisation that hires a fourth engineer has outgrown its plan without changing its code.
The file lifecycle is shorter than your product’s. Mubert track URLs expire and the docs tell you to persist what you need; SOUNDRAW download links die after 72 hours. Both mean the same architectural decision: generate, copy to your own storage immediately, and serve from there — which then collides with SOUNDRAW’s daily-refresh clause and with Mubert’s “no distribution” label. Resolve that in writing before the storage bucket is built, not after.
The licence changes shape at the moment of monetisation. SOUNDRAW requires negotiation before commercial use (section 3.6) and reserves a revenue-share discussion (3.6.1); Mubert’s grid flips “monetized content” on at Startup. A free beta can sit inside one set of terms and a paid launch in another, so the go-live checklist needs a licensing line item, not just a billing one.
Who should pick which
Mubert API fits products where music is delivered continuously rather than exported: fitness and focus apps, ambient or in-store layers, live channels, games that need a stream that never repeats, and teams that want quota and seat limits readable at runtime. The self-serve checkout, published plan grid and webhook support mean a two-engineer team can integrate without a sales cycle. Game teams weighing this route against hand-built soundtracks can compare it with the tools indie game developers actually ship with. It does not fit products that hand audio files to users on a subscription — that is a custom contract, and the published plans say so explicitly.
SOUNDRAW API fits creator tools where the output is a video the user exports: editors, social templates, presentation tools, UGC platforms. The section 4.4 transformation test is written for exactly that case, the entry price is the lowest of the two, and the application grant in 2.2 is broad enough to cover multi-user generation. It fits less well when you need a per-user seat model, published rate limits or an uptime commitment — none of those exist publicly, and the agreement is as-is with liability capped at twelve months of fees. Apps that use generated music as a calm background layer rather than an export — meditation and guided-session products, for instance — sit awkwardly between the two: the Beatoven.ai and Mubert comparison for guided meditation apps covers that delivery shape from the licensing side.
If neither shape dominates — for example, an app that both streams background music and lets users export clips — the honest answer is that you are buying two licences, and the export side will cost more than the streaming side on either vendor.
What we could not verify
Four gaps are real and are stated as gaps. Neither vendor publishes a service-level agreement with an uptime figure; SOUNDRAW’s agreement is explicit that the API is provided “as is” and “as available”, and no equivalent Mubert statement was retrievable, so uptime behaviour is unverified for both. Neither publishes latency or throughput benchmarks; Mubert’s sub-second WebRTC claim is marketing copy, not a measured figure. SOUNDRAW’s technical documentation, rate limits and per-user provisioning model are delivered after contracting, so the integration effort for a large multi-tenant product is unverified from public sources. And Mubert’s master API licence text is not published, which means the plan grid is the only public evidence of what sub-licensing and distribution actually permit — the contract you sign may say more.
No track from either API was generated, integrated, load-tested or measured for this article, which is why no score is assigned. The clauses are the evidence; they are enough to choose a tier and a contract shape, not enough to rank one vendor’s audio above the other’s.
Before you sign
- Name the delivery shape in the contract’s own vocabulary: playback, file export, or video export.
- Confirm which tier carries the right you need, by clause, not by plan name — Mubert’s sub-licensing sits one tier above the one most teams buy.
- Ask for the file-retention rule in writing: Mubert’s URL expiry and SOUNDRAW’s 72-hour links both force a storage design.
- Get the seat model confirmed if you provision per end user; a new Mubert client key consumes a seat, and SOUNDRAW’s per-user limits are not public.
- Put the monetisation trigger on the launch checklist, since both contracts change terms when money starts flowing.
- Ask what survives termination, and archive the signed agreement next to the release it covers.
Sources and update time
All sources were read on October 19, 2026. Prices on both vendors’ pages are volatile — Mubert was running a countdown sale and SOUNDRAW’s FAQ contradicts its own price table — so treat every figure as dated and re-check at checkout.
- Mubert API — product scope, text-to-music and image-to-music, track lengths up to 25 minutes, WebRTC streaming, sub-licensing as a named capability
- Mubert API v3 documentation — credential pairs, licence object and counters, webhooks, track store durations, URL expiry, one-stream-per-client rule
- Mubert API plans — Trial, Startup and Startup+ allowances, sub-licensing only on Startup+, “No distribution” on every published tier
- SOUNDRAW API — Starter and Pro pricing and employee scope, multi-user FAQ, documentation-on-request
- SOUNDRAW licence summary — all-plans grant, permitted and prohibited uses, subscription-linked publication
- SOUNDRAW Terms of Service — Article 9.3 non-sublicensable grant, Article 10 prohibition on distributing generated music without consent
- SOUNDRAW API General Agreement and Licensing Terms — section 2.2 application grant, 3.5 sublicensing prohibition, 3.17 72-hour download links, 3.8 Extended License revenue shares, 8.1 fees, 13.2 termination and perpetual use
Last verified: October 19, 2026. Score withheld because neither API was integrated, load-tested or measured; the licensing and integration evidence above is what the decision rests on.