Cyanite
Cyanite is a German AI platform that tags, describes and searches professional music catalogues. It auto-generates genre, mood, BPM and instrument metadata, finds tracks by similarity or plain-language prompt, and flags AI-generated recordings.
What is Cyanite?
Cyanite is the music-understanding layer that sits behind catalogue search. It is built by elceedee UG, a company based in Mannheim, Germany, and has been running since 2019. It does essentially one thing across three products: it listens to recordings and turns them into structured, searchable information.
Auto-Tagging analyses a track from beginning to end and returns some twenty-six categories of metadata — genres and subgenres, moods, BPM, tempo, key, instruments, vocal characteristics, character and movement, energy levels and emotional profiles, musical era and duration. Auto-Descriptions adds a natural-language summary of each song. Two search modes sit on top of that index: Similarity Search, which finds tracks resembling a reference recording by sound, feel and era, and Free Text Search, launched in 2023 as the first prompt-based music search engine, which converts a written description into catalogue matches. The 2025 Advanced Search release added percentage match scores, the strongest segment within each result, pre-filtering on the customer's own metadata, queries mixing library tracks with Spotify IDs, up to 500 results per query, and prompt translation from any language into English. A third product, AI Music Detection, arrived in 2026 and identifies material generated by Suno, Udio, Lyria, Mureka and ElevenLabs, deliberately calibrated to minimise false positives and unaffected by watermark-removal tools.
The models are developed in-house, with no reliance on external LLMs — a point the company makes repeatedly. The roots are academic: the founding team spun out of the Popakademie Baden-Württemberg, and the algorithms were developed alongside Music Information Retrieval research at TU Berlin.
Access comes three ways: the web app, a documented API, or native activation inside a music catalogue management system already in use. For catalogues that cannot leave the building, Cyanite offers a spectrogram pipeline, where audio is converted locally in a Docker container, and a pseudo-on-premise container running inside the customer's own AWS account.
The company reports more than 200 corporate customers, 250,000 artists, 200,000 free web app users, 50 million tracks analysed and a team of fourteen.
What it does
- Tag an entire catalogue automatically with genre, mood, BPM, key, instruments and vocal characteristics
- Generate a natural-language description for every track
- Find tracks that sound like a chosen reference song
- Search a catalogue with a plain-language prompt instead of keywords
- Detect whether a recording was generated by an AI model
- Push AI tagging into an existing music CMS without development work
- Embed music analysis in a third-party product through the API
When to use Cyanite / When not to
A quick filter to help you decide if Cyanite is the right fit.
When to use Cyanite
- Music publishers answering sync briefs, who need to surface licensable tracks from a large catalogue in minutes rather than days
- Production music libraries and sync marketplaces embedding search and recommendation into their own platform
- Distributors and labels that must produce clean, consistent metadata across catalogues of very different origins
- Audio branding agencies that need mood and genre data to justify a music choice to a client
- Music-tech teams and developers integrating audio intelligence through an API instead of building their own models
When not to use Cyanite
- Anyone looking to create music: Cyanite analyses recordings, it never generates them
- Producers wanting to edit, master or separate audio stems, none of which the platform offers
- Solo artists with a handful of tracks, since similarity search only becomes meaningful above roughly a thousand songs
- Teams that need a published price before engaging, as no rate appears anywhere on the site
- Users who require a non-English interface or a mobile application, neither of which exists
How to use Cyanite
A typical end-to-end flow, from setup to results.
- Create an account on the Cyanite web app, which allows five tracks per month at no cost
- Upload your music: mp3 at any bitrate up to fifteen minutes, with wav, mp4 and FLAC accepted for large catalogues
- Let the analysis run, typically from a few seconds to under thirty seconds per track
- Review the generated tags and the natural-language description in the library view
- Search the catalogue by similarity to a reference track, or by typing a free-text prompt
- Refine with Advanced Search: pre-filter on your own metadata, combine several reference tracks, read the percentage match scores
- For a large catalogue, deposit it on Cyanite's S3 infrastructure or share a download link
- For a product integration, follow the API documentation; any free account is eligible and results come back as JSON
- For sensitive material, run the supplied Docker container locally so that only spectrograms leave your environment, or deploy the analyser inside your own AWS account
- For volume beyond the self-service tiers, contact the business team or complete the enquiry form to obtain a quote
Pros & Cons
Pros
- Deep specialisation in music, rather than a general-purpose model repurposed for audio
- Models developed in-house, with no dependency on external LLMs such as ChatGPT
- Three distinct privacy models, one of which never lets the audio leave the customer's environment
- EU hosting and a German publisher, with GDPR compliance claimed on the product itself
- Customer ownership of uploaded content written explicitly into the terms
- Native integrations with most of the music catalogue management systems on the market
- A permanent free tier with API access included, so an evaluation costs nothing
Cons
- No published prices: the pricing calculator computes server-side and self-service tiers are visible only after signing in
- Complete silence on whether customer data is used to train the models, and no opt-out is documented
- No Data Processing Agreement published or offered, which is unusual for a European B2B vendor
- Tags cannot be corrected inside the product, and private accounts cannot export from the library view
- The isolated on-premise deployment loses every search feature and keeps tagging only
- AI detection covers a fixed list of generators, so newer models fall outside it until the list is updated
- English-only interface and documentation, and no mobile application
Pricing & Plans
A permanent free plan is available: the web app allows five tracks per month at no cost, and a free account is already eligible for API access. Beyond that, Cyanite publishes no prices. Self-service subscriptions and single analysis credits exist inside the application, but their amounts are visible only after signing in, and the pricing calculator reachable from the pricing page computes its figures server-side without displaying any published rate. Enterprise pricing is quoted on request and combines an API subscription, usage-based analysis fees and, where relevant, search subscriptions, with volume discounts. No entry price and no currency can therefore be stated.
- five tracks per month
- with API access included
- self-service tier
- one level documented at twenty tracks per month
- with single analysis credits available separately
- API subscription
- usage-based analysis fees and optional search subscriptions
- quoted on request with volume discounts
- pseudo-on-premise deployment limited to tagging
- without the search features
- access granted to selected non-commercial and research projects on application
Data, GDPR & hosting
A consolidated view of how Cyanite handles your data.
GDPR overview
Implementation is concrete rather than declaratory. The publisher is established in Germany, so the GDPR applies directly and no Article 27 representative is required or named. The privacy policy identifies elceedee UG as controller, represented by its three managing directors, and cites the German Telemedia Act, the Federal Data Protection Act and, since 25 May 2018, the GDPR. Users are granted free access to their data, plus rectification, blocking, erasure and objection to marketing use; accounts are deleted on request. The product itself claims EU-based servers and full GDPR compliance, and Google is described as a processor within the meaning of Article 28. Two gaps remain: no Data Processing Agreement is published or offered, and the rights section directs requests to an address on a domain that is not the site's own.
Who owns the data?
The terms are explicit on this point. Section 2.6 states that all rights, title and interest in Customer Data remain the property of the customer, including copyright, trademark and trade secret rights; the publisher receives only a limited licence, valid for the term of the agreement, to process that material for the purpose of delivering the service. Section 2.5 keeps the software, its improvements and the documentation with elceedee. Two clauses tilt the other way: section 2.7 allows elceedee to build and commercially exploit aggregated data and analytics, provided they are not identifiable as Customer Data, and section 2.8 assigns any customer feedback outright to the publisher.
Reuse rights
Customers keep their catalogue and the metadata produced from it, and may reuse both without asking permission: business accounts can request a CSV export of their tags, and API users receive their results as JSON as part of the plan. Two practical limits apply. Results cannot be edited inside the product, so a tag judged wrong has to be reported to support rather than corrected directly, and export from the library view is closed to private accounts. On the publisher's side, audio files are stated never to be shared with third parties, while aggregated, non-identifiable data may be reused freely for analytics and marketing.
Data retention & training
Hosting summary
Cyanite states that all audio is processed on secure servers located in the European Union, in full compliance with the GDPR. Storage and processing run on AWS, contracted through Amazon Web Services EMEA SARL, registered at 38 Avenue John F. Kennedy, L-1855 Luxembourg. That Luxembourg address identifies the contracting AWS entity rather than a data centre: no specific hosting country is named anywhere on the site, so the European Union is the only region that can be asserted. Large catalogues are deposited directly on Cyanite's S3 infrastructure or supplied through a download link. Two options reduce the footprint further: the zero-audio pipeline converts recordings into spectrograms locally, inside a Docker container, so only those representations leave the customer's environment; and the pseudo-on-premise deployment runs the tagging engine entirely inside the customer's own AWS account, with no data transfer outside it. On the website itself, Google Analytics operates with IP anonymisation, addresses being truncated within the EU and EEA, with full transmission to a United States server described as exceptional.
Things to keep in mind
Risks and trade-offs to weigh before adopting Cyanite.
- The site never says whether uploaded music is used to train the models, and documents no opt-out; for unreleased or contractually sensitive catalogues, that silence deserves a written answer before any upload
- The terms allow the publisher to build and commercially exploit aggregated data and analytics derived from customer catalogues, provided they remain non-identifiable
- Accepting machine tags without review can quietly standardise a catalogue around the model's categories, eroding the curatorial judgement that made it distinctive
- Tags cannot be corrected inside the product, so an error propagates through every downstream search until support intervenes
- AI music detection remains probabilistic: even a conservative model leaves room for a wrong call, and using it to accuse an artist would be a serious misuse
- Performance claims — six times more licensing requests, fifty-five percent less search time, ninety-nine percent accuracy — come from the vendor and are not independently verified
- With no published price, budget exposure is only discoverable through a sales conversation, which makes comparison with alternatives harder
Setup & Integrations
Technical difficulty
Difficulty depends entirely on the route chosen. The web app requires nothing beyond creating an account and uploading files. Activating Cyanite inside a catalogue management system already in use is described as requiring no technical lift. The API is a conventional developer integration, with public documentation and JSON responses. The two privacy-oriented options raise the bar: the zero-audio pipeline means running a Docker container locally, and the pseudo-on-premise deployment means operating the analyser inside your own AWS infrastructure, which assumes cloud skills. Large catalogue ingestion is handled by Cyanite once the files are deposited.
Deployment
Integrations
Supported languages
Behind Cyanite
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Who is Cyanite built for?
What metadata does it generate?
Is there a free option?
What does it cost beyond the free tier?
Which file formats and lengths are accepted?
Can I use it without sending my audio?
Does it work with my catalogue management system?
What happens to my data?
Should you pick Cyanite?
Cyanite is a specialist, and that is its main argument. It does not try to be a general-purpose audio tool: it listens to music catalogues and makes them searchable, through automatic tagging, natural-language descriptions, similarity and prompt-based search, and, since 2026, detection of AI-generated recordings. The models are built in-house rather than assembled on top of a third-party LLM, which is consistent with the rest of its positioning.
That positioning is unusually well matched to the anxieties of its market. A publisher or a film production sitting on unreleased material does not simply want good tagging; it wants to know where the audio goes. Cyanite answers with three graduated options: EU-hosted processing under the GDPR, a spectrogram pipeline in which the audio never leaves the customer's machines, and a container running inside the customer's own AWS account. The terms back this up by leaving ownership of uploaded content with the customer.
Two reservations deserve weight. The first is commercial: no price is published anywhere, so no budget can be estimated without a sales conversation. The second is more substantive: the site never says whether customer material is used to train the models, and documents no way to opt out. For a vendor that makes data control its central promise, that silence is conspicuous, and it is the first question worth asking before uploading a catalogue. No Data Processing Agreement is published either.
For a music company with a real catalogue and a real search problem, Cyanite is a credible and mature option — active since 2019, integrated into most catalogue management systems, and honest about the limits of its own detection model. Get the training question answered in writing, and the rest holds up.
- Choosing a selection results in a full page refresh.
- Opens in a new window.