Credits
2hives voices are built with open speech models and recordings that their makers share openly. Thank you to everyone below.
The credits are given in English, in the words the licences ask for.
Voices and speech models
The fr_claire voice was trained on the SIWIS French Speech Synthesis Database by Pierre-Edouard Honnet, Alexandros Lazaridis, Philip N. Garner and Junichi Yamagishi (Idiap Research Institute), licensed under CC BY 4.0; modified.
Built on Qwen3-TTS by the Qwen team, Alibaba Cloud, licensed under the Apache License 2.0; modified (fine-tuned) by 2hives.
Kokoro-82M by hexgrad, licensed under the Apache License 2.0. Kokoro-82M includes SIWIS French (CC BY 4.0, Honnet et al. 2017) and Koniwa (CC BY 3.0) data.
Speech recordings used for training
Contains Multilingual LibriSpeech (Pratap et al., Interspeech 2020), CC BY 4.0, derived from LibriVox public-domain recordings.
Contains audio from the European Parliament (© European Union), via VoxPopuli (Meta AI, CC0).
Voice data: CSS10 (Park & Mulc, Interspeech 2019), from LibriVox public-domain recordings. With thanks to the LibriVox volunteers.
Voice data: Open Home Foundation / Nabu Casa voice-datasets (CC0).
Ambience sounds
Some of the ambience layers in the player's Mix panel are recordings shared under a Creative Commons Attribution licence:
"250730_055349_PH_Ocean_waves_at_White_Beach" by kevp888 (https://freesound.org/people/kevp888/sounds/829444/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), modified: trimmed, looped, levelled.
"strong wind river-side 121226_03.flac" by klankbeeld (https://freesound.org/people/klankbeeld/sounds/172522/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), modified: trimmed, looped, levelled.
"small running water forest stream_1.wav" by Kranhytt (https://freesound.org/people/Kranhytt/sounds/671883/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), modified: trimmed, looped, levelled.
"020A_100111_0243_EXP02_Bells_drone" by kevp888 (https://freesound.org/people/kevp888/sounds/844152/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), modified: trimmed, looped, levelled.
Where each voice comes from
| fr_claire | SIWIS French Speech Synthesis Database (CC BY 4.0) |
|---|---|
| fr_julien | CSS10 French, LibriVox public domain (CC0) |
| hu_anna, hu_berta, hu_imre | OHF voice-datasets (CC0) |
| hu_csilla | CSS10 Hungarian, LibriVox public domain (CC0) |
| all fine-tuned voices | Qwen3-TTS base model (Apache-2.0); training data also from Multilingual LibriSpeech (CC BY 4.0) and VoxPopuli (CC0, European Parliament) |
We only use recordings whose licences allow it, and never recordings from video sites, podcasts, films or radio. info@2hives.com
