Credits & Attributions

The built-in "stock" voices you can choose on Mayantar (one female and one male voice for English, and one female and one male voice for Hindi) are generated by conditioning our text-to-speech models on short reference recordings taken from the openly licensed speech corpora credited below. This page gives the attribution those licences require.

• Work: Indic TTS Hindi speech database, created by the Speech Lab, Indian Institute of Technology Madras, and the TTS Consortium of Indian institutions, funded by the Ministry of Electronics and Information Technology (MeitY), Government of India.
• Source: https://www.iitm.ac.in/donlab/indictts/database (we obtained the Hindi recordings through the Hugging Face mirror https://huggingface.co/datasets/SPRINGLab/IndicTTS-Hindi).
• Licence: Creative Commons Attribution 4.0 International (CC BY 4.0) — https://creativecommons.org/licenses/by/4.0/

• Work: "LibriTTS-R: A Restored Multi-Speaker Text-to-Speech Corpus" (Yuma Koizumi and colleagues, Google, Interspeech 2023), a speech-restored version of LibriTTS. The underlying readings are by volunteer readers for the LibriVox project.
• Source: https://www.openslr.org/141/ (obtained through https://huggingface.co/datasets/mythicinfinity/libritts_r).
• Licence: Creative Commons Attribution 4.0 International (CC BY 4.0) — https://creativecommons.org/licenses/by/4.0/

We selected short passages from these recordings and modified them: some were joined end to end, all were cut to about ten to eleven seconds with a short fade-out, resampled to 24 kHz mono and level-normalised. The resulting clips are used only as reference audio for voice synthesis. The audio you hear when you generate speech is newly synthesised by our models, is watermarked as AI-generated, and is not a recording of the original speaker saying your text.

The creators of these corpora and the speakers in them do not endorse Mayantar or its Service, and are not affiliated with it. The licences above apply to the original datasets; nothing on this page grants you any rights in the speakers' voices beyond those licences.

If you are a creator or contributor and believe a credit here is incomplete or incorrect, contact us at legal@bimbient.com and we will correct it promptly.