Licences / Licences

Letterboard

Copyright © 2026 the Letterboard authors. Licensed under the GNU General Public License, version 3 (full text). The web version's own code is published unminified on this site (index.html, app.js, boards.js, mathtext.js, piper.js, style.css, sw.js). To request the source of the Android app, write to support@letterboard.app.

Built-in speech engine (web version)

The built-in voices run in the browser with the following components, served from /engine/.

eSpeak NG – GNU GPL v3 or later

Pronunciation rules and language data (piper_phonemize.wasm, piper_phonemize.data). Version: the rhasspy/espeak-ng fork at commit 8593723f10cfd9befd50de447f14bf0a9d2a14a4.

• Source code: espeak-ng-8593723f….tar.gz (16 MB, hosted here) · upstream github.com/rhasspy/espeak-ng
• Licence: GNU GPL v3 (some data files carry other compatible licences; see the source archive)

piper-phonemize – MIT

Converts text to the phonemes the voices use; compiled together with eSpeak NG into piper_phonemize.wasm and piper_phonemize.js. Version: the wide-video/piper-phonemize fork at commit cfff8e52ebaea37c7e953ae2d06b174acb827ac4. Copyright © 2023 Michael Hansen.

• Source code: piper-phonemize-cfff8e52….tar.gz (9.8 MB, hosted here) · upstream github.com/wide-video/piper-phonemize
• Licence: MIT

piper-wasm build – MIT

The compiled files come from the npm package @diffusionstudio/piper-wasm 1.0.0 (Konstantin Paulus; declared MIT), built with Emscripten 3.1.47 from the two sources above using the published steps: build instructions · github.com/diffusion-studio/piper-wasm (based on wide-video/piper-wasm).

Emscripten runtime – MIT / University of Illinois NCSA

The JavaScript glue in piper_phonemize.js is generated by Emscripten 3.1.47. Licence text.

ONNX Runtime Web 1.20.1 – MIT

Runs the voice models (ort.wasm.min.js, ort-wasm-simd-threaded.mjs, ort-wasm-simd-threaded.wasm), from the npm package onnxruntime-web 1.20.1. Copyright © Microsoft Corporation.

• Licence: MIT · third-party notices for software included in ONNX Runtime: ThirdPartyNotices.txt
• Source code: github.com/microsoft/onnxruntime (v1.20.1)

Built-in voices

Downloaded on request from Hugging Face. Each voice has its own licence and attribution below.

• Aivars (lv_LV-aivars-medium): rhasspy/piper-voices, CC0 1.0.
• Rudolfs (lv_LV-rudolfs-medium): RaivisDejus/Piper-lv_LV-Rudolfs-medium, CC0 1.0. Made by Raivis Dejus from audiobooks of Latvijas Neredzīgo bibliotēka.
• Mari — Estonian (speaker 2 of et_EE-news-medium): model card. Trained on the Speech Corpus of Estonian News Sentences, distributed through the University of Tartu under CC BY (META-SHARE “Available — Unrestricted Use”); fine-tuned from LibriTTS-R.
• Reginutė — Lithuanian (lt_LT-reginute1-medium): published by RobertasTa under CC BY 4.0. Trained on the LIEPA corpus (2013–2015), Vilnius University (Institute of Mathematics and Informatics; Faculty of Philology), with the Institute of the Lithuanian Language, Lithuanian University of Educational Sciences and Šiauliai University. Project lead: prof. Laimutis Telksnys; corpus work and documentation: Gediminas Navickas. Published to Hugging Face by MEL DYNAMICS, MB as meldynamics/liepa-tts. The publisher discloses an unresolved licensing question in the Irina → Lessac base-model lineage; see the linked model documentation.
• Lithuanian pronunciation data and code: stress dictionary derived from Arūnas Smaliukas' svogunas/g2p-lt-lexicon and LIEPA annotations, CC BY 4.0. Pronunciation and number-expansion code by RobertasTa, GPL-3.0, adapted to JavaScript for Letterboard. Pinned original sources and adaptation notes; JavaScript source. Voice files and downloaded dictionaries are used without modification.

Android app

• Kotlin standard library 2.0.21 (JetBrains) and JetBrains annotations: Apache License 2.0 (full text). Source: github.com/JetBrains/kotlin.
• Speech uses the text-to-speech engines installed on the device, which are separate apps with their own licences.