Credits, attributions and licenses

Lire cette page en français

The Monk & The Witch is a free fan project with no monetization by Fred Potvin. Its books and website draw on the work of software developers, speech researchers, voice-data contributors and type designers. This page acknowledges those contributions and links to the terms published by their rights holders.

For audio-component selection, we apply a commercial-use compatibility policy, even though the project earns no revenue. This screening policy does not change the free fan-content terms described below.

The audiobook voices are generated by AI text-to-speech systems. Fred selects the voices, directs their use and prepares the audio edition. The synthetic narration is not a recording of the credited researchers or source speakers performing these books.

Credits reviewed: 14 September 2026. This register covers the complete installed English and French voice library, including all speakers in the 23 Piper models, all 29 installed Kokoro presets, all ten Supertonic 3 styles in both languages, and the configured NeuTTS French models and supplied references. These credits apply to current and future narration, dialogue and samples made with this library, independently of character casting. Choosing another covered speaker or changing pace, gain or effects does not require new credits. A newly added model, source or license revision does require a coverage update.

Credit here does not mean that every voice appears in a published chapter.

• Fan content and original work
• How the audiobook is made
• Speech engines and models
• English Piper voice credits
• French Piper voice credits
• Understanding the license references
• Fonts, icons and website tools
• Corrections and attribution requests

Fan content and original work

The saga uses elements of Dungeons & Dragons and the Forgotten Realms. The Wizards of the Coast Fan Content Policy governs the project’s use of Wizards’ material. The full unofficial-fan-content notice appears in the footer of this page and throughout the fiction site.

Wizards’ material remains the property of its respective rights holders. The project’s original contributions and use of generative AI are described on the author’s rights page. These credits do not imply sponsorship, approval or endorsement by Wizards, the software developers, the dataset contributors or the source speakers.

The license links below identify particular third-party resources. They do not place the entire saga, its artwork or its audio edition under a single software or Creative Commons license, and they do not replace any rights or conditions that apply to those resources.

How the audiobook is made

Speech is synthesized locally using downloaded models. The website plays prepared audio files; listeners do not need to install the speech engines or download model weights.

The production process assigns voices to narrative passages and dialogue, adjusts pronunciation and speaking rate, assembles the passages with pauses, and balances their volume. Some character voices also receive pitch or echo effects. These are production choices made for this edition. They should not be attributed to the people who supplied the training recordings.

Published website audio uses the Ogg container and Opus codec. The production tools include FFmpeg for assembly, effects and encoding, Opus for audio compression, and ONNX Runtime for running the speech models. We thank their developers and contributors. Their respective references are FFmpeg licensing, Opus licensing and the ONNX Runtime MIT license.

The local pipeline also uses Misaki by hexgrad and contributors for pronunciation processing (Apache-2.0), spaCy by Explosion and contributors for English text analysis (MIT), PyTorch by its contributors for neural processing (license and contributor notices), and SoundFile by Bastian Bechtold and contributors for audio I/O (BSD license), backed by libsndfile (LGPL). We also acknowledge the eSpeak NG and Phonemizer pronunciation tools and the llama.cpp runtimes linked with NeuTTS below. These are software credits for local production.

Speech engines and models

Kokoro v1.0

Kokoro-82M v1.0, by hexgrad and contributors, provides the installed American English, British English and French presets. The model and bundled voice presets use Apache License 2.0. Our ONNX distribution comes from kokoro-onnx, by thewh1teagle and contributors; its runtime code is MIT licensed.

This credit covers every installed preset in the official voice inventory:

• American English: af_heart, af_alloy, af_aoede, af_bella, af_jessica, af_kore, af_nicole, af_nova, af_river, af_sarah, af_sky, am_adam, am_echo, am_eric, am_fenrir, am_liam, am_michael, am_onyx, am_puck, am_santa.
• British English: bf_alice, bf_emma, bf_isabella, bf_lily, bm_daniel, bm_fable, bm_george, bm_lewis.
• French: ff_siwis.

The French ff_siwis preset additionally credits the SIWIS French Speech Synthesis Database (2017): Junichi Yamagishi, Pierre-Edouard Honnet, Philip Garner and Alexandros Lazaridis, University of Edinburgh, Centre for Speech Technology Research, DOI 10.7488/ds/1705, CC BY 4.0. This is the source-dataset attribution requested by Kokoro; the model’s Apache license is a separate layer. We use the supplied models for new synthetic speech, with pronunciation, timing, level and optional effects adjusted for the books.

The model card also acknowledges @rzvzn for training, Li et al. for StyleTTS 2, and the iSTFTNet research. Its shared training-data credits include Koniwa contributors and Television Nishinippon (TNC) for the Koniwa tnc recordings, under CC BY 3.0; see the source collection. These are model-provenance acknowledgements, not additional installed voices.

Supertonic 3

Supertonic is developed by Supertone Inc. The production configuration uses Supertonic 3, covering styles F1, F2, F3, F4, F5, M1, M2, M3, M4 and M5 in English and French, with the OpenRAIL-M model license. The Python package is separately MIT licensed.

The model license requires a clear machine-generated-content disclosure. This page and the footer identify the audiobook voices accordingly. See the model license at the revision used and the Python package license.

NeuTTS Nano French

The local voice configuration also includes NeuTTS Nano French Q4 and Q8, developed by Neuphonic Limited, for prepared French narration and auditions. These configurations are not evidence that a NeuTTS chapter has been published. The backbones use the NeuTTS Open License v1.0, which permits commercial use of models and outputs within its revenue limit; this solo project earns no revenue. See the recorded licenses for Q4 and Q8.

The NeuCodec INT8 ONNX decoder is separately Apache-2.0. The configured Juliette reference is supplied by Neuphonic in the official reference collection, alongside Dave, Emily, Greta, Jo, Mateo, Paul, Sophie and Steven. The NeuTTS package and supplied references use the NeuTTS Open License v1.0 at this revision, including its commercial revenue condition. The voice names identify synthetic reference presets, not performers of these books.

Local NeuTTS processing also uses eSpeak NG (GPLv3), Phonemizer (GPLv3), llama.cpp (MIT) and llama-cpp-python, by Andrei Betlen and contributors (MIT). We thank their authors and contributors; their software is distinct from the generated narration.

Piper

Piper is maintained by the Open Home Foundation and contributors, building on the earlier work of Michael Hansen and Rhasspy. The installed production package, piper-tts 1.8.0, declares GPL-3.0-or-later. The current engine’s license differs from the MIT license of the older Rhasspy engine.

Piper’s engine and its downloadable voice models have separate provenance. Each model credit below links to its original card in the Rhasspy Piper voice collection. The license labels reproduce the source terms identified by those cards; most appear under the card’s Dataset heading. They are not a claim that Piper provides one uniform license for every voice or for every generated recording.

English Piper voice credits

One entry covers each model, regardless of how many characters use it. For both multi-speaker LibriTTS and LibriTTS-R models, this acknowledges the corpus team and the underlying LibriTTS, LibriSpeech and LibriVox contributions. Source reader identities are provenance, not audiobook performer credits. The internal voice catalog also uses LibriTTS-P, by Masaya Kawamura, Ryuichi Yamamoto, Yuma Shirahata, Takuya Hasumi and Kentaro Tachibana, for speaker and speaking-style annotations under CC BY 4.0; selection and presentation of those annotations are project metadata processing.

French Piper voice credits

The installed French library includes all 125 MLS speakers, Siwis, and both UPMC speakers, alongside the Kokoro, Supertonic and NeuTTS voices credited above. UPMC’s Jessica and Pierre speakers belong to one multi-speaker model. The original corpora are documented separately in the Jessica repository and the Pierre repository.

Understanding the license references

The engine code, model weights, training recordings and generated audiobook are different materials. A license applying to one does not, by its name alone, establish all rights in the others. For example, GPL version 3, section 2 addresses when program output falls within the GPL. Creative Commons likewise explains that license conditions depend on the rights exercised by a particular use.

The main labels in the register mean:

The project review is settled for the recorded models and this prepared-audio workflow. We accept reasonable uncertainties in the relationship between training data, model weights and newly synthesized narration, while excluding direct sources with incompatible noncommercial, research-only or reserved-rights terms. Required credits and applicable conditions remain in force. This is a project-use decision, not a license grant to other users or a waiver of third-party rights.

Fonts, icons and website tools

The website serves the following typefaces. Copies of their supplied copyright notices and complete SIL Open Font License 1.1 texts are available with the font files:

Interface icons use Font Awesome Free 7.3.0, by Fonticons, Inc., delivered through cdnjs. Its webfonts use OFL 1.1 and its CSS uses MIT; its SVG/JavaScript icon distribution uses CC BY 4.0. See the Font Awesome Free license. Brand icons identify their respective services and remain subject to their owners’ trademark rights.

The site is generated with Hugo, by the Hugo Authors, under Apache License 2.0. We also acknowledge the libraries and contributors supporting these tools. This page covers the principal reader-facing resources and audiobook technologies; it is not a redistributable software bundle or an inventory of every development dependency.

Corrections and attribution requests

For a missing credit, a correction to a source or license, or a question about a specific recording, contact fred@monkandwitch.com. Please identify the chapter or resource and, where possible, link to the relevant original notice.

Production records retain model identifiers, source-card snapshots and recording metadata.