HERTZKIT / INFORMATION
Credits
Authors, projects, models and licenses of components included in HertzKit. Follow the links to their provenance and full notices.
Updated: 11 November 2026.
On this page5 sections
Third-party libraries, models and data
GrooVAE / Magenta.js
MIDI Groove Lab uses the pretrained Magenta groovae_2bar_humanize neural model, Magenta.js 1.23.1 and TensorFlow.js 2.7.0. They are served locally for inference in a Worker. Library code and model weights are unmodified. The interface, playback synthesis, examples, result blending and MIDI utilities are HertzKit implementations.
GrooVAE — Jon Gillick, Adam Roberts, Jesse Engel · Magenta.js · Apache 2.0 · TensorFlow.js — Apache 2.0
Training reference: Groove MIDI Dataset — Google LLC, CC BY 4.0. No recordings or MIDI files from the dataset are redistributed. Reference: Gillick, Roberts, Engel, Eck and Bamman, “Learning to Groove with Inverse Sequence Transformations”, ICML 2019. Use of the model does not imply Google or Magenta endorsement of HertzKit.
Notices and licenses for bundled components: protobufjs · core-js · regenerator · long · seedrandom · Provenance and file checksums
LAME / lamejs
The converter’s MP3 output uses @breezystack/lamejs 1.2.7, a JavaScript adaptation of LAME distributed under LGPL-3.0. The package credits Alex Zhukov and preserves other author notices in its sources. HertzKit distributes the library unmodified, separately from the converter interface.
LAME · lamejs · LGPLv3 · GPLv3 · Notices and replacement instructions · Corresponding source
This component’s licenses retain their rights and conditions, including permitted modification and debugging. Their use does not imply endorsement of HertzKit by the authors.
Basic Pitch
AI model for audio-to-MIDI transcription, developed by Spotify. Apache 2.0 license. github.com/spotify/basic-pitch
HertzKit local adaptation: transcription runs in a cancelable Worker and inference tensors are released between batches. Original notices and licenses are retained. Export uses fixed-pitch notes.
TensorFlow.js
In-browser machine learning engine, developed by Google, used to run the Basic Pitch model. Apache 2.0 license.
Tone.js MIDI
Library for reading and writing MIDI files in JavaScript. MIT license. github.com/Tonejs/Midi
CMU Pronouncing Dictionary
English pronunciation dictionary (125,000+ words), developed by Carnegie Mellon University, used in the rhyme dictionary. BSD license. speech.cs.cmu.edu/cgi-bin/cmudict
FrequencyWords (hermitdave)
Spanish word frequency list, compiled by Hermit Dave from subtitles in the OpenSubtitles corpus, used to expand the rhyme dictionary's database. The project's code is MIT and its word lists (the content) are under the CC BY-SA 4.0 license. github.com/hermitdave/FrequencyWords
The rhyme dictionary's Spanish word database (rhymes-es-full.json) is an adaptation of that list: it keeps the words and their frequency order, without the counts, and adds the rhyme keys HertzKit calculated for each one. For that reason the file is shared under the same license, CC BY-SA 4.0, and you may reuse it with that attribution and under those terms.
Separator model and licenses
UVR-MDX-NET-Voc_FT · Ultimate Vocal Remover; web-audio-separation · Dan Kurtz; ONNX Runtime Web · Microsoft; PFFFT · Julien Pommier, Hayati Ayguen, NCAR FFTPACK. Separator model and licenses.
Score engraving and PDF
Verovio 5.6.0 · LGPL-3.0-or-later · Verovio. Unmodified modules, loaded separately. GPL · LGPL · Corresponding source.
Fonts included in Verovio: Leipzig, Bravura, Gootville, Petaluma, Leland and Liberation · SIL OFL 1.1. Leipzig · Bravura · Gootville · Petaluma · Leland · Liberation
jsPDF 4.2.1 · MIT · jsPDF · License. PDF uses 300 dpi page images. Distribution notices.
MIDI-event to MusicXML conversion and original examples were developed for HertzKit. Engraving libraries do not detect notes from audio.
Region data
Country and region choices: subdivision data from iso-codes distributed with pycountry 26.2.16 (LGPL-2.1). The account uses top-level entries and localized names, with manual entry available. License · Source.
Recorded instruments and examples
Listening studies: EQ, compression, stereo and drum placement, with CC0 audio and documented settings. Provenance and permission
Music for video, podcasts and games: six CC0 examples for comparing arrangement, endings and loops, with listening tasks and a delivery worksheet.
Reference piano and practice example
VCSL recordings by Versilian Studios and contributors, distributed under CC0 1.0. HertzKit selected and trimmed the samples, prepared a playback bank and arranged the practice example. The example waveform is generated from that audio.
Recorded classical guitar
Spanish classical guitar, FreePats project. Source version 2019-06-18, CC0 1.0. HertzKit prepares selected recordings for browser playback, balancing levels and compensating small tuning deviations.
FreePats · Spanish classical guitar · CC0 1.0 · Sample provenance
Upright bass and percussion
Upright bass recordings are from VSCO 2 Community Edition; percussion is from VCSL. Sam Gossner and contributors (Versilian Studios), CC0 1.0. The banks retain sample provenance, licenses and hashes.
Original HertzKit development
HertzKit develops the interfaces, musical examples and original utilities used by its tools. Third-party components and their licenses are listed on this page.
Audio, tuning and practice
The tuner, metronome, trainers, Practice and audio review tools include interfaces, reference sounds and features developed for HertzKit. Original musical examples are generated without commercial recordings or artist voices.
HertzKit Practice
The practice journey, take comparison, suggested tasks and reports are HertzKit developments. Tasks are selected using local rules. PDF export uses jsPDF, credited above.
Musical resources
Original chord diagrams, patterns, melodies and synthesized examples were created for the tools. Measurements are working aids, without certification or endorsement by ITU, EBU or other organizations cited as references.
MIDI and lyric tools
MIDI Workshop, Lyric and rhythm, the progression editor and their examples are HertzKit developments. Syllable splitting reuses its local rhyme engine. External dictionaries and components retain the licenses and attribution listed on this page.
Vocal pitch estimation
The vocal trainer uses an original HertzKit implementation of YIN’s cumulative difference method with local filtering and resampling. Scientific reference: Alain de Cheveigné and Hideki Kawahara, “YIN, a fundamental frequency estimator for speech and music”, Journal of the Acoustical Society of America 111(4), 1917–1930 (2002). No article code or content is redistributed. The exercises and detection engine are HertzKit’s own work. Playback uses the banks credited under “Recorded instruments and examples”.
Guides and examples
Guide texts and original examples were prepared for HertzKit. References identify their authors and sources; naming a service does not imply endorsement or affiliation.
Written examples in the AI music workshop are original HertzKit drafts. Suno and Udio functions are explained with source links; their models, interfaces and songs are not redistributed. Example usage permission.
HertzKit Sessions
Activity format, validation, archiving, interface and reader component: HertzKit development. Reuses the original Practice, MIDI and audio engines and the recorded banks already credited on this page. Music examples are original HertzKit exercises. The exported reference uses a 16-bit mono WAV copy; it does not include the full sound bank.
SHA-256 checks audio integrity. It is not a signature, timestamp or authorship certificate. Format and reader documentation: Sessions 1.1.
The integration demonstration reuses the recorded piano samples and original teaching arrangement credited above. Its two example takes are assembled for teaching; they are not student recordings. The listening-decision worksheet and worked text example were prepared for HertzKit.
Music-work examples and worksheets
The work-guide examples, worksheets and performance-budget interface are original HertzKit material. Sources are identified in the relevant guides and do not imply affiliation or endorsement. The calculator shows its arithmetic and does not calculate taxes, establish rights or certify agreements.
Fonts and license notices
Fonts
Fraunces (The Fraunces Project Authors), IBM Plex Sans and IBM Plex Mono (IBM Corp.), under the SIL Open Font License 1.1. They are served from this same site, not through Google Fonts. The text of each license: Fraunces, IBM Plex Sans, IBM Plex Mono.
Full license texts
The complete texts of the licenses for the CMU Pronouncing Dictionary, TensorFlow.js, Basic Pitch and Tone.js MIDI, and the notice about FrequencyWords, are in the third-party license notices file.
External services
Hosting, messages and measurement
The site is hosted on Cloudflare. Visit measurement through Google Tag Manager uses Google services. Cloudflare Turnstile verifies contact form submissions, and Resend delivers the messages. What data each one processes is explained in the privacy policy and the cookie policy.
Newsletter — Brevo
Brevo manages the optional HertzKit email subscription forms, confirmations and campaigns. See Privacy for data processing and subscription choices.
Accounts and security
Cloudflare provides hosting and Turnstile security checks. Resend delivers sign-in codes. The account interface and optional project saving were developed for HertzKit. Data handling is explained in Privacy and Cookies; musical components retain their licenses.
Advertising and affiliate links
When enabled, Adsterra supplies the optional banner on eligible public pages. Its content comes from the advertising network, separately from HertzKit’s tools and affiliate links. See the privacy and cookie notices for your choices.
CJ and other affiliate networks may provide links, external banner images and optional impression pixels. Their use does not imply an endorsement of HertzKit. Permission and data handling are explained in Privacy and Cookies.