Rightsify retired its Hydra generator for a data-licensing product

The audio task

A producer or business needing quick, rights-cleared instrumental music, and a separate AI developer needing large licensed datasets for training, have different problems. Rightsify's product history shows one company answering both, under different names, at different times – which matters for anyone citing "Hydra" today.

What the documents show

Rightsify's own December 2023 launch post describes Hydra as a generative model "trained on a dataset of 60,000 songs from Rightsify's library," sold as a $99-a-month subscription for businesses and creators to generate instrumental tracks, samples and sound effects for their own commercial use – not as a training-data licence for other AI companies. An archived January 2025 snapshot of the Hydra product page states consumer access ended "as of September 1st 2024" and that Hydra remained available to businesses only. As retrieved on 16 September 2026, Rightsify's live Gramosynth page no longer markets Hydra at all; it instead offers AI labs a "training data pipeline" built on Rightsify's "world's largest licensed music corpus," matching the third-party-licensing role the topic of AI training data is usually associated with. The three documents describe three different moments in one company's roadmap, not one stable product.

Rights status

Rightsify's 2023 launch post states generated Hydra output was "backed by Rightsify's copyrights, granting full freedom for commercial use." Neither the 2023 post nor the 2025 archive states that Hydra itself licensed source material to other AI developers; that training-data-licensing role is what the current Gramosynth page describes. None of the three documents discloses which specific catalogues, artists or labels are included in the underlying corpus, so the scope of what is actually cleared cannot be verified beyond Rightsify's own description.

What to check before you use it

This is an editorial checklist. Confirm which Rightsify product a source is actually describing before citing it, note that Hydra as a consumer generation tool is discontinued, and ask Rightsify directly what recordings and composers sit inside the Gramosynth corpus before treating any output or licence as fully cleared.

• Am I citing the discontinued Hydra generator, or the current Gramosynth training-data product?
• Does Rightsify disclose which catalogues sit inside the corpus I would be relying on?
• What happened to output already generated by subscribers before the September 2024 consumer cutoff?

A platform's own naming can move faster than the record of what it actually built, which is why the retrieval date on a living product page matters as much as the page itself.