Sync Licensing Companies 2026: Models, Splits and AI Policy

Sync licensing companies all promise the same outcome, which is your music underneath somebody's pictures. What separates them is the business model sitting underneath the promise: whether they take the track exclusively, whether they buy it outright, what share of the fee reaches you, and whether the same track can earn anywhere else while it sits on their platform.

That distinction matters more than any ranked list of sync licensing companies. For how a sync licence is actually cleared — the two separate clearances every placement needs, and why controlling both sides of the copyright is the thing that closes deals — our sync licensing guide is the mechanism page. This one stays on the companies: what each model really costs, what the placements pay, what gets a catalogue accepted, and where AI-generated music stands with each type.

Our interest in this is narrower than a buyer's guide. We test whether files survive automated gates, and a sync submission has one that most artists never see. Before a curator at any of these companies presses play, the upload has passed through metadata validation, duplicate detection and, increasingly, automated provenance checks. An AI catalogue can be rejected at that layer without a human ever forming an opinion about the music.

Three business models, and what each one really costs

Almost every company in this market is one of three things wearing different branding. Work out which one you are looking at and most of the contract becomes predictable.

The exclusive library or buyout. You assign rights — sometimes just the master, often the composition too — in exchange for a fixed upfront payment, a share of downstream royalties, or both. Epidemic Sound is the clearest example, with reporting through early 2026 describing a fixed buyout in the low thousands per accepted track plus a 50/50 split on streaming royalties the platform generates. The money is the most predictable in this market. The cost is that the track stops being yours to license, and if it becomes valuable later, that value is not yours either.

The non-exclusive marketplace. You upload, you keep your rights, the platform takes a commission on each licence sold. Songtradr, Pond5 and Audiosocket sit here. Per-licence fees are the lowest of the three models, frequently in the tens or low hundreds, and the compensation is volume plus the freedom to have the identical track earning on four platforms at once. This is where almost everyone should start, and where a lot of people should stay.

The boutique agency. A small curated roster, hands-on pitching, roughly 50/50 on placement fees that can reach five figures, and an acceptance rate that makes the other two models look welcoming. Musicbed and Marmoset are the reference points, both built around narrative indie and cinematic material rather than production-library utility. The cost here is not a percentage. It is that you probably cannot get in, and that reading their terms is a poor use of an afternoon until you can.

There is a fourth thing that looks like a company and is not: the paid-membership pitching service. Music Gateway is the best-known, charging a subscription for access to briefs and opportunities. Some artists get placements from it. The structural point is that you are paying for access rather than being paid for results, which reverses the usual incentive, so treat the monthly fee as marketing spend and judge it on returns.

The companies compared

Two caveats on that table, and they are not decoration. Splits and tier pricing in this market change without announcement, and several of these companies restructured during 2025 and 2026. And a percentage means nothing until you know what it is a percentage of: 80% of a $60 marketplace licence is worse than 50% of a $9,000 agency placement, and both are worse than 100% of a direct deal you found yourself.

Songtradr in one paragraph

Songtradr is the marketplace most independent artists encounter first, and its tiering is the part that confuses people. The entry tier costs around $19 a year and the platform's commission on it is high, in the region of 60%, while the Pro tier at roughly $49 a year cuts that to about 20%. Published figures have moved more than once, and the arithmetic only works out in Pro's favour above a certain annual earnings threshold. Our Songtradr review works through the tiers, the commission maths and the AI position in detail, so we will not duplicate it here.

What placements actually pay, by tier

Fee ranges get quoted as if they were rate cards. They are not. A sync fee is negotiated against a production's music budget, which means the same track is worth $150 to a podcast and $30,000 to a car advert. These are the tiers as we see them reported and quoted in 2026, and they are indicative rather than promised.

The honest reading of that table is the access column rather than the money column. Everything above roughly $5,000 is gated by representation rather than by quality, because a network campaign will not carry a chain-of-title risk on a stranger regardless of how the track sounds. The bottom two rows are genuinely open, and they are where an independent catalogue with depth, edits and clean metadata makes actual money.

There is also back-end. Television and film placements generate performance royalties through cue sheets, collected by your PRO, and on a broadcast placement that back-end can exceed the upfront fee over time. That is only true if the composition is registrable, which is exactly where the AI question stops being philosophical — our ASCAP vs BMI comparison covers the registration side and which society to join.

Reading the contract: exclusivity, term, territory, MFN

Four clauses decide whether an offer is good, and none of them is the percentage everyone fixates on.

Exclusivity is the scope of what you cannot do elsewhere. Check whether it covers the master only or the publishing too, and whether it applies to the specific recording or to the composition in any form. An agreement that takes both sides of the copyright leaves you nothing to license anywhere.

Term is how long that lasts. Two years with a renewal you have to actively agree to is a reasonable ask. Perpetual is not a synonym for long, it means forever, and a perpetual exclusive on a track you generated last Tuesday is a permanent decision about an asset you have not evaluated yet. Look for a reversion clause and read what triggers it.

Territory is where the exclusivity applies. Worldwide is the standard ask and usually fine, but it is worth noticing when an offer is worldwide on exclusivity and narrower on where they can actually place you.

MFN, most favoured nations, ties your terms to somebody else's. In a fee negotiation it means the master and publishing sides are quoted at matching rates so neither outruns the other. In a library agreement it can guarantee terms no worse than any other contributor. If you own both sides of your own copyright, MFN in a fee context is mostly a formality — one of the quiet advantages of a single-owner catalogue.

The rule of thumb we would apply: a two-year non-exclusive at 30% beats a perpetual worldwide exclusive at 50% every time, unless the exclusive comes with a named campaign attached.

What a library checks before it accepts a catalogue

Rejections at this stage are overwhelmingly procedural. Curators are assessing whether you are a reliable supplier, not whether the song moved them.

One clear owner on master and publishing. A placement needs both cleared, and libraries have learned that chasing a third co-writer through a picture deadline is how deals die. A solo catalogue where one person controls both sides clears in a single signature, and that is a genuine commercial advantage rather than a consolation.

Instrumental versions and stems. Full track, instrumental, and cuts at 60, 30 and 15 seconds. Stems where you can produce them, because a buyer who can mute a lead line will take the track that lets them do it over the one that does not.

Broadcast-ready masters. This is the delivery detail that catches streaming-first artists out. Broadcast television is mixed far quieter than streaming, around -24 LUFS against Spotify's -14 and Apple's -16, and a master slammed for streaming loudness arrives at a library already wrong. Our AI music mastering guide covers the targets and why one master does not serve both markets.

Honest metadata. Mood, tempo, key, instrumentation, use case, vocals or not. Supervisors search by brief, and inflated or vague tagging gets a catalogue quietly deprioritised rather than removed. Your ISRC codes belong here too, since cue sheets need them.

An indemnity you can stand behind. Every one of these agreements asks you to warrant that you own what you are licensing and to cover the buyer's costs if that is untrue. If you cannot honestly make that warranty — an uncleared sample, an undocumented collaborator, a model licence you never read — that is a contract to walk away from, not a form to tick.

The AI question, stated honestly

We are not going to publish a per-company table of AI policies, because it would be wrong within weeks. Several of these companies changed position during 2025 and 2026, some quietly, and the only authoritative source is the submission agreement in front of you today.

What we can describe is the pattern, which follows the business model rather than the brand. Non-exclusive marketplaces are the most permissive, because they carry the least warranty exposure per track and pass a good deal of it back to the uploader. Buyout and subscription libraries have the strongest reason to restrict AI material, since they are warranting a clean catalogue to thousands of subscribers simultaneously and one contaminated track is a systemic problem rather than a single refund. Boutique agencies rarely need a written ban, because curation for narrative indie material excludes most generated output before policy is relevant.

Submission forms increasingly ask the question directly, and some go further and ask what the model was trained on. Answer both honestly. A library that discovers the truth later removes your entire catalogue, not the one track, and the removal is the mild outcome compared with an indemnity claim. Registration is a separate constraint worth planning around: ASCAP, BMI and SOCAN aligned in October 2025 to accept partially AI-generated compositions while holding fully AI-generated ones ineligible, which directly affects the cue-sheet back-end described above. Our Suno copyright explainer covers the ownership position, and we are not lawyers.

The gate before the gate

Here is the part specific to an AI catalogue, and the reason this page exists on a site that benchmarks detection rather than reviews libraries. Before any of the human judgement above happens, the file itself gets read by machines.

Every generated track leaves its tool carrying passengers: a SynthID-class watermark woven into the signal, a C2PA provenance manifest naming the model, a spectral fingerprint unique to the generator, and secondary layers underneath. They are inaudible by design. They are also exactly what automated screening reads, which is why our AI music detector explainer matters as much to a sync submission as to a distributor upload. A catalogue can be flagged at ingestion and never reach curation at all.

Undetectr is the first and only AI music watermark remover — built specifically to remove what platform screening scans for rather than adapted from a repair suite. It clears all six artifact layers in one pass, runs in the browser with nothing to install, takes under a minute per track, accepts MP3, WAV and FLAC from Suno, Udio and ElevenLabs Music, and masters to each platform's loudness spec in the same pass. SoundMatch checks for fingerprint collisions before release, which matters more here than anywhere else, because a collision discovered after a campaign airs is a categorically worse problem than one discovered on Spotify. It is €39 once for unlimited tracks, with a €19 Starter tier at 10 credits, and outputs for 150+ platforms.

Across our 50-file AI music corpus, processed files passed 49 of 50 production distributor classifiers, with the single failure an unrelated copyright flag on a remix. Untreated files passed none at DistroKid's threshold. The manual alternatives we tested were worse and slower: iZotope RX 11 cleared 32 of 50 at $399 and four to six hours a track, Audacity 8 of 50. Our watermark remover benchmark has the full method, and our Undetectr review is the sceptical long-form assessment. Both carry the same caveat: a remover will not fix a weak song, will not get anyone to hear it, and classifiers get retrained.

Then the framing that keeps people out of trouble. Clearing artifacts is release prep. It is never a substitute for disclosing AI use where a contract asks. Those obligations are independent. A clean file does not make a false warranty true, and Undetectr is not a defence against a misrepresentation claim. Run the file so it is technically clean, answer the AI question honestly, and pitch the companies that accept AI-assisted material. That combination is a working business.

The honest verdict

If you are starting, start non-exclusive. Songtradr, Audiosocket and Pond5 will take open submissions, the fees are small, and nothing you sign there prevents you from signing something better later. Build to 30 or 50 usable tracks with edits, stems and real metadata before approaching anyone curated, because libraries buy suppliers rather than songs.

Treat exclusivity as something you sell rather than something you concede, and buyouts as a genuine option rather than a trap — a fixed few thousand for a track that would otherwise earn $40 a year is a rational trade, provided you understand it is permanent. The boutique agencies pay best and accept least, and the tiers above $5,000 stay closed to unrepresented catalogue however good the music is.

And whichever model you pick, the file has to clear the machines before it reaches the people. That is the one item on this page Undetectr handles for you. Everything else is paperwork you have to do yourself.

Where to go next: our sync licensing guide for how the two clearances work, streaming royalties for what the same catalogue earns on the other side of the business, how to make money with AI music for the wider income picture, and our methodology for how we test everything on this site.