Legal
Module B: Audio and voice
This is the full text of the current Module B: Audio and voice. It is published, not click-accepted: a purchase records the version in effect at the sale by its SHA-256, and the Order License of an order attaches it. You can request an earlier version from support@graunt.com.
sha256:d5e5b3aff0c4c73af858d2c4f409e430207681852bb87788bb027810ec2a7b63
# Graunt — Marketplace License, Module B: Audio and voice **Version:** 1.0 | **Effective:** October 31, 2026 | **Status:** Operational terms. **Governs:** Audio recordings, speech, music and ambient sound, and the soundtrack of video. Part of the Marketplace License (graunt.com/legal/marketplace-license) for content of this kind. --- ## 1. What this module covers This module applies to every listing whose content is audio, and to the soundtrack of video. It adds to the Marketplace License; where it is more specific, it governs for this content. Words have the meanings the Marketplace License gives them. ## 2. What you may do with audio Where the listing allows model training, you may decode, resample, segment, transcribe, align, label, embed and annotate the audio and use all of it, in whole or in part, for the purposes §2 of the Marketplace License names, including training speech recognition, audio understanding, text-to-speech and audio generation models. Transcripts, alignments, embeddings and other derived data you make are yours, except to the extent they reproduce a recognizable portion of the content; a transcript of speech in the content is the content. ## 3. Voices Where the content records people's voices, you do not use it to clone a person's voice, to synthesize a particular speaker's voice, or to make a model or feature that reproduces a particular speaker; you do not use it to identify a speaker, to build speaker recognition, or to create voiceprints or other biometric identifiers; and you do not label the content with the names or identities of the people speaking. Training a general speech or audio model on many voices is not any of these. A text-to-speech or voice model trained on the content may not be tuned, prompted or offered to reproduce a particular speaker in it. Nor do you use the content to build, train or deploy a system that sorts speakers by race or ethnicity, health, sex life or sexual orientation, religion or beliefs, political opinions or union membership from their voice, or that reads a particular speaker's emotions to monitor or assess them, and you do not try to contact, locate, track or profile anyone recorded. The release status a listing declares for the people recorded, what a release covers and records, and what you do when a person withdraws consent, are as the video and people-on-camera module and §7 of the Marketplace License describe. The personal-information module applies to speech that holds personal information. ## 4. Music and recordings A listing states, as its seller declares, the clearance of music and audio in the content. Music and recordings are licensed for the purposes the listing allows and for no media use: you do not publish, stream, distribute or publicly perform them, or any recognizable portion of them, and you do not use a model to output them, unless the listing grants public use and then only to the extent it names. A model that outputs a recognizable recording or composition from the content reproduces it, and §2 of the Marketplace License says what follows. Where a listing grants public use of music, any public-performance or other license a collecting society requires for that use is yours to obtain, unless the listing says the music carries no such fees, and you file any cue sheets that society's rules require. You do not register the music or the recordings, or claim them in a content-identification system, and you do not collect money for them, as §10 of the Marketplace License says. ## 5. No media use The Marketplace License grants no media use of audio unless the listing grants public use. Your staff and contractors may listen to the content to prepare, label, select and evaluate it. --- ## Version History | Version | Date | Changes | |---------|------|---------| | 1.0 | October 31, 2026 | Initial version, published with Marketplace License 2.0: what a buyer may do with audio and the soundtrack of video; no cloning, synthesis or identification of a particular speaker, no voiceprints, no sorting speakers by sensitive traits or reading their emotions, and no contacting or tracking anyone recorded; releases and withdrawn consent as the video module and the License say; music and recordings, with collecting-society fees on public use and no registration or content-identification claims; no media use. |