Licences and notices
Speechable is built on open-source software and openly licensed models. This page lists those components and their licences, and provides the attributions their licences require. We are grateful to everyone who made them.
Speech model use restrictions
The Supertonic-3 model is licensed under the BigScience Open RAIL-M License, which places binding restrictions on how it may be used — including prohibitions on impersonating people without consent and on publishing generated content without disclosing that it is machine generated. These apply to you when you use Speechable's speech features. See the Text-to-Speech Model Terms and the full licence text.
Models
Machine-learning models used for speech synthesis, transcription and document understanding.
Cloud and on-device speech synthesis. Speechable distributes fp16 and int8 quantised derivatives. Use restrictions apply — see the Text-to-Speech Model Terms.
On-device speech synthesis (legacy eco mode) and hosted inference.
Software
Server-side image resizing and re-encoding, via sharp. Licence text, written source offer and relinking information are on the linked page.
Phonemisation for the on-device speech engine. Fetched by your browser directly from the jsDelivr CDN at runtime — Speechable does not host or redistribute it.
User-interface components.
Icons.
Document editor.
PDF parsing.
Web article extraction.
DOCX conversion.
Image processing.
Browser-support data compiled into the client bundle.
Fonts and artwork
Typeface.
Full dependency list
The components above are those whose licences require attribution, or that are central to how Speechable works. Speechable also depends on several hundred further open-source packages, predominantly under the MIT, ISC, Apache-2.0 and BSD licences. For a complete list, or for a copy of any licence not linked here, contact [email protected].