SermonWave uses open data and open-source software made by others. This page gives the credit their licences require, and some we give as thanks.
1. SermonWave voices
Voices whose names end in "(SermonWave)", such as "Mateo (SermonWave)" and "Lucía (SermonWave)", were built by SermonWave. The voices and the models behind them belong to SermonWave, as section 17 of our Terms of Service explains. The credits below are for the data and software we built them with. Each of those stays under its own licence, and nothing on this page changes the rights that licence gives.
2. Speech data used to train SermonWave voices
| Data | Licence | Credit |
|---|---|---|
| Google crowdsourced Latin American Spanish speech data sets Argentina (openslr.org/61), Chile (71), Colombia (72), Peru (73), Puerto Rico (74), Venezuela (75) |
CC BY-SA 4.0 | Contains speech data from the Google crowdsourced Latin American Spanish data sets, Copyright 2018, 2019 Google, Inc., licensed under CC BY-SA 4.0. Described in A. Guevara-Rukoz, I. Demirsahin, F. He, S.-H. C. Chu, S. Sarin, K. Pipatsrisawat, A. Gutkin, A. Butryna and O. Kjartansson, "Crowdsourcing Latin American Spanish for Low-Resource Text-to-Speech", LREC 2020. Changed by SermonWave: we selected recordings, trimmed silence, resampled them, levelled their volume and normalised the text before training. |
| Multilingual LibriSpeech (MLS), Spanish openslr.org/94 |
CC BY 4.0 | Contains speech data from Multilingual LibriSpeech by V. Pratap, Q. Xu, A. Sriram, G. Synnaeve and R. Collobert, "MLS: A Large-Scale Multilingual Dataset for Speech Research", Interspeech 2020, licensed under CC BY 4.0. The recordings come from public-domain LibriVox audiobooks; thank you to the LibriVox volunteers. Changed by SermonWave: we selected some speakers and recordings, resampled them, levelled their volume, and re-transcribed the text with punctuation. |
| Mozilla Common Voice, Spanish commonvoice.mozilla.org |
CC0 1.0 | No credit is required. We used it in earlier training stages and thank the Common Voice contributors. |
| Sermon audio from churches that chose to help | Permission under section 5 of our Terms | Some SermonWave voices may also be trained on sermon audio from organisations that have agreed to it. Each voice is blended from many speakers. We do not name these churches or speakers, to protect their privacy. |
All of this data is provided by its licensors "as is", without warranties. See each licence for its full terms, including its disclaimer of warranties.
3. Starting models
Our Spanish voices start from the openly published Piper voice model es_ES davefx (medium) from the Piper project's training checkpoints (rhasspy/piper-checkpoints). Its model card lists its data set (OHF-Voice voice-datasets) under CC0. We then trained it further on the data above.
4. Open-source software
| Software | Licence | What we use it for |
|---|---|---|
| Piper (OHF-Voice/piper1-gpl; earlier versions rhasspy/piper) | GPL-3.0 (earlier versions: MIT) | Training voices and turning text into speech |
| eSpeak NG (espeak-ng/espeak-ng) | GPL-3.0 or later | Turning written text into the sounds a voice should speak |
| ONNX Runtime (microsoft/onnxruntime) | MIT | Running voice models |
| PyTorch and Lightning (pytorch/pytorch, Lightning-AI/pytorch-lightning) | BSD-3-Clause, Apache-2.0 | Training voices on our own servers |
Open-source licences keep applying to that software wherever it runs, including on SermonWave devices. Where we run GPL software on a device, you can get its source code from the links above, or by emailing team@serve.love.
5. Voices made by others
Voices without "(SermonWave)" in their name were made by others. We do not own them. Each stays under its own licence.
| Voice | Licence | Credit |
|---|---|---|
| Upc Ona (ca_ES, Piper, medium) Source |
CC BY-SA 3.0 ES | UPC Festcat TTS corpora (Universitat Politècnica de Catalunya), CC BY-SA 3.0 ES |
| Upc Ona (ca_ES, Piper, x_low) Source |
CC BY-SA 3.0 ES | UPC Festcat TTS corpora (Universitat Politècnica de Catalunya), CC BY-SA 3.0 ES |
| Upc Pau (ca_ES, Piper, x_low) Source |
CC BY-SA 3.0 ES | UPC Festcat TTS corpora (Universitat Politècnica de Catalunya), CC BY-SA 3.0 ES |
| Bu Tts (cy_GB, Piper, medium) Source |
CC BY 4.0 | bu-tts Welsh/English dataset (Techiaith, Bangor University), CC BY 4.0 |
| Mls (de_DE, Piper, medium) Source |
CC BY 4.0 | Multilingual LibriSpeech (Pratap et al., Meta AI; OpenSLR 94), CC BY 4.0 |
| Alba (en_GB, Piper, medium) Source |
CC BY 4.0 | Alba voice dataset (University of Edinburgh DataShare 10283/3270), CC BY 4.0 |
| Aru (en_GB, Piper, medium) Source |
CC BY 4.0 | ARU speech corpus (Acoustics Research Unit, University of Liverpool), CC BY 4.0 |
| Jenny Dioco (en_GB, Piper, medium) Source |
Jenny (Dioco) dataset licence: commercial use permitted with attribution | Jenny (Dioco). Voice dataset by Dioco (github.com/dioco-group/jenny-tts-dataset); commercial use with attribution |
| Northern English Male (en_GB, Piper, medium) Source |
CC BY-SA 4.0 | Google crowdsourced UK and Ireland English dialect speech (OpenSLR 83), CC BY-SA 4.0 |
| Southern English Female (en_GB, Piper, low) Source |
CC BY-SA 4.0 | Google crowdsourced UK and Ireland English dialect speech (OpenSLR 83), CC BY-SA 4.0 |
| Vctk (en_GB, Piper, medium) Source |
CC BY 4.0 | CSTR VCTK Corpus (Centre for Speech Technology Research, University of Edinburgh), CC BY 4.0 |
| Libritts (en_US, Piper, high) Source |
CC BY 4.0 | LibriTTS (Zen et al., Google; OpenSLR 60), CC BY 4.0 |
| Libritts R (en_US, Piper, medium) Source |
CC BY 4.0 | LibriTTS-R (Koizumi et al., Google; OpenSLR 141), CC BY 4.0 |
| Antton (eu_ES, Piper, medium) Source |
CC BY 4.0 | Antton Basque dataset (itzune), CC BY 4.0 |
| Maider (eu_ES, Piper, medium) Source |
CC BY 4.0 | Maider Basque dataset (itzune), CC BY 4.0 |
| Mls (fr_FR, Piper, medium) Source |
CC BY 4.0 | Multilingual LibriSpeech (Pratap et al., Meta AI; OpenSLR 94), CC BY 4.0 |
| Mls 1840 (fr_FR, Piper, low) Source |
CC BY 4.0 | Multilingual LibriSpeech (Pratap et al., Meta AI; OpenSLR 94), CC BY 4.0 |
| Siwis (fr_FR, Piper, low) Source |
CC BY 4.0 | SIWIS French Speech Synthesis Database (SIWIS project, University of Edinburgh), CC BY 4.0 |
| Siwis (fr_FR, Piper, medium) Source |
CC BY 4.0 | SIWIS French Speech Synthesis Database (SIWIS project, University of Edinburgh), CC BY 4.0 |
| Upmc (fr_FR, Piper, medium) Source |
CC BY-SA 4.0 | UPMC Pierre voice data (MaryTTS, github.com/marytts/upmc-pierre-data), CC BY-SA 4.0 |
| Iseke (kk_KZ, Piper, x_low) Source |
CC BY 4.0 | KazakhTTS (ISSAI, Nazarbayev University), CC BY 4.0 |
| Issai (kk_KZ, Piper, high) Source |
CC BY 4.0 | KazakhTTS (ISSAI, Nazarbayev University), CC BY 4.0 |
| Raya (kk_KZ, Piper, x_low) Source |
CC BY 4.0 | KazakhTTS (ISSAI, Nazarbayev University), CC BY 4.0 |
| Google (ne_NP, Piper, medium) Source |
CC BY-SA 4.0 | Google crowdsourced Nepali speech (OpenSLR 43), CC BY-SA 4.0 |
| Google (ne_NP, Piper, x_low) Source |
CC BY-SA 4.0 | Google crowdsourced Nepali speech (OpenSLR 43), CC BY-SA 4.0 |
| Mls (nl_NL, Piper, medium) Source |
CC BY 4.0 | Multilingual LibriSpeech (Pratap et al., Meta AI; OpenSLR 94), CC BY 4.0 |
| Mls 5809 (nl_NL, Piper, low) Source |
CC BY 4.0 | Multilingual LibriSpeech (Pratap et al., Meta AI; OpenSLR 94), CC BY 4.0 |
| Mls 7432 (nl_NL, Piper, low) Source |
CC BY 4.0 | Multilingual LibriSpeech (Pratap et al., Meta AI; OpenSLR 94), CC BY 4.0 |
| Mls 6892 (pl_PL, Piper, low) Source |
CC BY 4.0 | Multilingual LibriSpeech (Pratap et al., Meta AI; OpenSLR 94), CC BY 4.0 |
| Edresson (pt_BR, Piper, low) Source |
CC BY 4.0 | TTS-Portuguese Corpus (Edresson Casanova et al.), CC BY 4.0 |
| Artur (sl_SI, Piper, medium) Source |
CC BY 4.0 | Artur studio TTS (ppisljar, Hugging Face), CC BY 4.0 |
| Alma (sv_SE, Piper, medium) Source |
CC BY 4.0 | NST Swedish speech synthesis data (Språkbanken, National Library of Norway); model CC BY 4.0 |
| Padmavathi (te_IN, Piper, medium) Source |
CC BY 4.0 | IndicVoices-R (AI4Bharat, IIT Madras), CC BY 4.0 |
| Venkatesh (te_IN, Piper, medium) Source |
CC BY 4.0 | IndicVoices-R (AI4Bharat, IIT Madras), CC BY 4.0 |
| Vais1000 (vi_VN, Piper, medium) Source |
CC BY 4.0 | VAIS-1000 Vietnamese speech synthesis corpus (VAIS), CC BY 4.0 |
| Google (bn_BD, Piper, medium) Source |
CC BY-SA 4.0 | Google crowdsourced Bengali speech (OpenSLR 37), CC BY-SA 4.0, and the CMU Indic voices (Carnegie Mellon University) |
| Spanish training data: Google crowdsourced Latin-American and Argentinian Spanish (OpenSLR 61, 71, 72, 73, 74, 75) Source |
CC BY-SA 4.0 | Guevara-Rukoz et al., "Crowdsourcing Latin American Spanish for Low-Resource Text-to-Speech" (LREC 2020), Google; used to train our Spanish voices Mateo and Lucía |
| Spanish training data: Multilingual LibriSpeech, Spanish (OpenSLR 94) Source |
CC BY 4.0 | Pratap et al., "MLS: A Large-Scale Multilingual Dataset for Speech Research" (Interspeech 2020), Meta AI; LibriVox recordings; used to train our Spanish voices Mateo and Lucía |
6. Questions
If you think a credit here is missing or wrong, email team@serve.love and we will fix it.