Cyanite.ai

What is Cyanite.ai?

Cyanite.ai is an AI-based music tagging and search platform operated by the German company elceedee UG; it helps music publishers, copyright libraries, synchronization licensing teams, distributors, streaming services, and creators to organize and discover tracks.

After uploading audio, users can obtain structured music metadata, natural-language descriptions, and search vectors, which allow them to find matching content in their own music collection by referring to existing tracks or textual descriptions.

Main functions

• Automatic music tagging: It analyzes an entire audio file and generates fields such as genre, sub-genre, mood, BPM, speed, key, instruments, vocal characteristics, era, energy level, and appropriate usage scenarios.
• Automatic description: It transforms musical characteristics into natural-language descriptions, enabling copyright-owned music libraries to explain the sound, atmosphere, and potential uses to customers.
• Similar track search: Starting from a reference track, it identifies similar pieces in one’s own music collection based on sound, atmosphere, and era; filtering can also be applied using criteria such as genre, BPM, or key.
• Natural language search: Enter details such as movie scenes, events, emotions, cultural cues, or specific music requirements, and the system will match these semantics with the audio files in its database; it does not rely on external chat models to carry out the search.
• Advanced search: It allows combining multiple reference tracks, as well as using custom metadata for filtering and sorting; it returns similarity scores, the most matching segments, and up to 500 results. This feature is primarily intended for technical teams through APIs.
• AI music detection: It identifies music generated by common creation models, and is used for content review, ensuring transparency and compliance; the range of music that can be detected depends on the models currently supported.
• Music emotion curve: The API provides information on how the emotion of a song changes second by second, which is useful for identifying specific segments, analyzing the progression of the narrative, or matching videos with particular emotional moments.
• Enterprise customization: Business clients can negotiate customized tags, classification systems, bulk directory import, CSV outputs, and dedicated testing environments.

Input and output

Web usage process

1. Register an account and access the music library; a free account allows for the analysis of 5 tracks per month.
2. Upload MP3 files that meet the requirements; for large-scale directories or those in other formats, it is necessary to arrange their import in coordination with the business team.
3. Wait for the system to complete the analysis, and check the genre, mood, BPM, key, instruments, vocals, and automatic description.
4. Select a reference track to perform a similar search, or enter specific textual details regarding the scene, rhythm, mood, and purpose.
5. Use existing metadata, BPM, genre, similarity, and other criteria to narrow down the results and listen to the most relevant segments.
6. The selected results can be used for organizing music libraries, making recommendations, or identifying candidates for licensing; however, key tags and copyright decisions should still be reviewed by professionals.

Price and quota

Cyanite offers free tiers and self-service points for individuals, while for businesses it provides pricing based on API subscriptions as well as combinations of usage analysis and search capabilities. The public website does not list all the costs associated with self-service subscriptions in detail.

290 euros is the basic fee for using the public API; it does not cover all costs related to analysis and searching. Factors such as region, taxes, the size of the catalog, bulk discounts, and the duration of the contract can affect the final price.

API and developer integration

Cyanite offers a GraphQL API that allows developers to upload files, initiate analyses, retrieve structured results, and use similar search, natural language search, and advanced search functions.

1. Create a free account, and in the integration settings generate an Access Token, a Webhook address, and a Webhook Secret.
2. Use Bearer authentication to submit files or playlist objects to the GraphQL interface, thereby triggering the necessary audio analysis.
3. Verify the Webhook signature and receive events indicating success or failure, then query detailed results using the analysis ID.
4. Map genres, moods, BPM, key, instruments, second-by-second mood, and other fields to your own database.
5. Based on business requirements, similar search or text search can be utilized, along with custom metadata filtering, control over the number of results, sorting options, and multiple reference tracks.
6. Tokens are managed separately for testing and production, errors, quotas, and version changes are logged, and it is prevented from exposing credentials in the frontend code.

Official GitHub and open-source status

The official GitHub organization provides examples of JavaScript integrations, examples from AWS Marketplace, and starter projects for music hackathons; some of these examples are licensed under the MIT license.

These repositories are used to demonstrate the API or deployment process, and they cannot be used as evidence that Cyanite’s music analysis models, search algorithms, or the Web App itself are open source; the older integration example pages also do not specify any license details.

Platforms and integration

Three privacy processing modes

Copyright, data, and contract considerations

• The customer retains the right to its Customer Data and grants the platform permission to process it as necessary for providing the services; the customer must still ensure that the music, metadata, and manner of use are legal.
• The service provider can create aggregated data and analyses based on customer data that prevent the identification of individual customers, and retain such aggregated results; organizations that are sensitive to training data or its secondary use should clarify the scope through contracts.
• Deleting a Cyanite account will remove personal data and uploaded audio, but exceptions may apply for communications, legal obligations, investigations, and backups.
• Free web services have a different contract structure from corporate orders. Unless otherwise specified, the initial duration for corporate orders is usually one year, with automatic renewal; even if cancellation occurs ahead of time, the remaining subscription fee may still be required to be paid.
• Upon termination of the service by the company, the terms require that customer data be destroyed; if a copy of the data is needed, it must be requested in writing at least 14 days in advance, and the cost of transferring the data may apply.
• Automatic tagging, search results, and AI music detection can only assist in catalog management and review; they cannot replace the verification of rights chains, copyright assessments, or human oversight of music.

Suitable for users and typical use cases

• Music publishers and music library creators: Standardize old catalog labels and enable quick search for candidates suitable for synchronization use in film and television scenes.
• Publishers and streaming services: Generate metadata in bulk, and establish capabilities for recommendations, playlist creation, and content moderation.
• Music supervisor and brand team: They use reference tracks or descriptive statements in natural language to find pieces that suit the visuals, the mood, and the brand’s tone.
• Artists, producers, and DJs: Understand the characteristics of their own tracks, create descriptions for them, and discover similar music.
• Music technology developers: Integrate analysis, search, AI-based music detection, and recommendations into their products through APIs.

Capabilities boundaries and limitations

• The platform does not allow direct editing of analysis results at the moment, and users of private pages cannot export the complete tags directly; any errors or missing classifications require contact with support or adjustment in an external system.
• Similarity search always returns the tracks that are most similar, but even in a small music library the best results may not be similar; the platform recommends that a catalog of over 1,000 tracks will enable this feature to work better.
• The more specific the natural language prompts are, the more controllable the results tend to be; vague terms do not ensure a unique or subjectively correct match.
• Large-scale catalogs, formats other than MP3, CSV exports, custom categorization, and advanced privacy settings all require business or technical support.
• The functions of different privacy modes are not identical; isolated AWS deployments do not support all three search capabilities.

Frequently Asked Questions

How many songs can Cyanite.ai analyze for free each month?

The free Web App offers 5 analyses per month. Free accounts can also be used to set up API integrations, but the amount of testing available is limited.

What are the requirements for uploading audio?

On the web interface, it is possible to upload MP3 files of any bitrate with a duration of up to 15 minutes. WAV, MP4, FLAC files, as well as large directories, require batch import.

Can the label results be modified or exported on the web page?

Currently, it is not possible to edit the results within Cyanite, and private users cannot export the Library View directly; commercial clients can request CSV or table format results.

Does the 290 euro fee for the API include all costs?

It is not included. This amount represents the public monthly fee for using the API; the total cost also depends on the size of the music library, the volume of analysis performed, search subscriptions, and the combination of features used.

Will the uploaded concerts be processed by third-party models?

The standard search and tagging algorithms were developed by Cyanite; the cloud processing guidelines state that audio data is not shared with third parties. For projects that require a high level of sensitivity, it is possible to use a spectrogram pipeline or an isolated AWS deployment.

Is Cyanite.ai an open-source product?

It is not an open-source product that has been officially recognized. The vendor has made some APIs and deployment examples available publicly; some of these are licensed under the MIT license, but the core platform and models are not made open source as a result.