Whisper model weights by OpenAI; Q5_1 conversion distributed by Georgi Gerganov. MIT.
whisper.cpp and ggml by Georgi Gerganov and contributors. MIT.
Kotlin and AndroidX: Apache License 2.0. LLVM libc++: Apache 2.0 with LLVM exceptions.
Inter by Rasmus Andersson and the Inter Project Authors. SIL Open Font License 1.1.
Unmodified InterVariable.ttf from https://github.com/rsms/inter/releases/tag/v4.1

Keyboard suggestion filter: List of Dirty, Naughty, Obscene, and Otherwise Bad Words
(LDNOOBW), © 2012–2020 Shutterstock, Inc., and contributors. English, Spanish, German
and Norwegian lists. Source: https://github.com/LDNOOBW/List-of-Dirty-Naughty-Obscene-and-Otherwise-Bad-Words/tree/5faf2ba42d7b1c0977169ec3611df25a3c08eb13
Creative Commons Attribution 4.0 International: https://creativecommons.org/licenses/by/4.0/
Full licence: cc-by-4.0.txt. Hex lowercased and classified the entries, omitted ordinary,
medical and search-only terms, and added common slurs, inflections and spelling variants.

English, Spanish and German keyboard language models (.hextm): derived from Google Books Ngram
Viewer, version 3 (2020-02-17), by Google and the Google Books Ngram contributors.
https://storage.googleapis.com/books/ngrams/books/datasetsv3.html
Creative Commons Attribution 3.0 Unported: https://creativecommons.org/licenses/by/3.0/
Changes: lowercased and summed word and word-pair counts for 1980–2019, filtered
spellings, selected informative pairs, and quantized probabilities into HXTM v3.

German uses the German corpus's alphabetic spellings; no external German dictionary.
English literal validity also uses SCOWL words, proper names, uppercase forms and abbreviations
outside the frequency model, lowercased and sorted; these spellings carry no word probabilities.
English spellings: SCOWL 2020.12.07, sizes through 70; Kevin Atkinson and contributors.
Permission and individual source notices: scowl.txt. https://wordlist.aspell.net/
Spanish spellings: RLA-ES Hunspell dictionaries (es_ES and es_MX), coordinated by Santiago Bosio,
used under Mozilla Public License 1.1. The frequency model filters by these spellings;
the original dictionaries are unmodified. See mpl-1.1.txt, rla-LICENSE.md and rla-README_hunspell_es.txt.
Source files: https://github.com/LibreOffice/dictionaries/tree/762abe74008b94b2ff06db6f4024b59a8254c467/es

Norwegian Bokmål keyboard language model: N-grammer fra Nasjonalbiblioteket 2026,
National Library of Norway (Nasjonalbiblioteket). Book unigrams and bigrams ("digibok"),
language nob, 1980–2019. CC0 1.0 Universal: https://creativecommons.org/publicdomain/zero/1.0/
https://data.norge.no/nb/datasets/f2eebe8b-bd83-34e4-bcf3-c0c5e37a77f8/n-grammer-fra-nasjonalbiblioteket-2026
Spelling filter: Norsk ordbank – bokmål 2005, 1 February 2022, University of Bergen
(Universitetet i Bergen) and the Language Council of Norway (Språkrådet), distributed by
Nasjonalbiblioteket. CC BY 4.0: https://creativecommons.org/licenses/by/4.0/
https://data.norge.no/en/datasets/f0db926d-a154-3a12-82b8-6573b57f6dbb/norsk-ordbank-norwegian-bokmal-2005
Changes: normalized and lowercased word forms, filtered the book counts to attested
Bokmål spellings, selected informative pairs and quantized probabilities into HXTM v3.
