Illustrations, word and example recordings, and adapted dictionary data: Word Atlas contributors — CC BY 4.0
Source: https://github.com/designfuchen/word-atlas
License: https://creativecommons.org/licenses/by/4.0/
Changes: all 785 A1, 1039 A2, 1465 B1, 2127 B2, 1608 C1, 274 C2 entries selected; original 36 short examples and action scenes retained. Added 7262 compositional meaning diagrams, five-second transport, synchronized word recordings and gentle synthesized event sounds, level browsing and practice. The source AI-generated data and programmatically composed diagrams have not been proofread entry by entry. Not an official or exhaustive CEFR word list.
Media adaptation: illustrations optimized with a 384-pixel WebP profile; speech optimized as mono 24 kbit/s MP3 at 22.05 kHz. Original older media is retained in source history, and originals for this update are retained in the local evidence backup. Previously optimized B1/B2 media is unchanged.

Typing normalization and insertion metrics adapted from Word Atlas src/lib/typing.ts, commit 99c007673d38f02c418d8e4cd2e9ec90b5d40be5 (MIT). Copyright and license: licenses/word-atlas-MIT.txt. Independent practice and mistake-book integration implemented for WordMotion.

Ambient recordings: Word Atlas public/sfx/amb-{rain,sea,forest,fire,voyage,cafe}.mp3, pinned commit 99c007673d38f02c418d8e4cd2e9ec90b5d40be5, repository MIT license (licenses/word-atlas-MIT.txt). Six original files retained byte for byte. The reading room, theme controls, route and Canvas galaxy integration are implemented for WordMotion.
Classic English prose: Lewis Carroll, Alice's Adventures in Wonderland (1865), Project Gutenberg #11; H. G. Wells, The Time Machine (1895), Project Gutenberg #35. Sources and complete original texts including the Gutenberg license are linked from each book. Chinese introductions and the six bilingual short stories are original WordMotion learning material, AI-assisted, not official exam passages. Book prose is English; introductions are not complete translations.

Fun Studio: 24 original AI-assisted bilingual cultural learning cards, with native Unicode symbols and CSS artwork. Terminology references: Royal Museums Greenwich https://www.rmg.co.uk/collections/objects/rmgc-object-14575 ; IAU https://www.iau.org/IAU/Astronomy-FAQs/Constellations.aspx ; Victoria and Albert Museum https://www.vam.ac.uk/articles/the-lunar-zodiac . No artwork or source prose copied. Existing animal recordings retain the Word Atlas attribution above; other names and examples use device speech synthesis. Fun cards are supplementary topics, not additions to the CEFR source count.

Version 10: 1011 second-example recordings from Word Atlas public/atlas/audio/ex2/, pinned source commit above, CC BY 4.0. Sentences are matched against the existing second examples, source Git blob hashes verified, recordings normalized to mono 24 kbit/s MP3 at 22050 Hz and fully decoded. Original source files are retained outside the deployment. Other second examples use device TTS with the interface labeled accordingly.
FSRS-6 scheduler: official ts-fsrs 5.4.2, MIT; unmodified ESM build in vendor/ts-fsrs-5.4.2.js and license in licenses/ts-fsrs-MIT.txt. Integrity and source reference: source/ts-fsrs-vendor.json. WordMotion migration, legacy shadow, learning modes, insights, offline cache policy, and card export are original integration work. Default target retention is 90%; no personal parameter fitting is claimed.
Reading expansion: six additional original bilingual A1-C2 suggested-level shorts and two original cultural-topic shorts (14 original shorts total), AI-assisted. Cultural references for the two topic shorts are linked in the texts. Device TTS is used for reading, not recorded narration. These are not official CEFR assessments or the reference site's full online reading corpus.


WordMotion v12 exam preparation (2026-10-30)
36 original short method drills, passages, scripts, answer explanations and sample responses were written for WordMotion with AI assistance. They are not official IELTS/ETS questions, endorsed courses, complete mock examinations or calibrated scores. Format and scoring facts link to IELTS/ETS primary sources within the exam area. Audio scripts use optional device speech synthesis; local recordings and written responses remain on the learner's device. Vocabulary illustrations and word recordings retain the Word Atlas attribution above.


WordMotion v13 bilingual static discovery (2026-10-30)
English headwords, licensed source senses, illustrations, original examples and recordings retain their attribution above. Traditional-script pages are derived writing conversions; original English text and recorded media remain unchanged. OpenCC JS 1.4.2 is redistributed with its MIT/Apache-2.0 and third-party notices in licenses/opencc-*. Static word pages add related vocabulary navigation and links to WordMotion learning cards. The new learning guides are original WordMotion material produced with AI assistance; source-derived vocabulary and script conversions have not received word-by-word professional review.

WordMotion v15 galaxy engine/layout adapt Word Atlas MIT code from src/galaxy/engine.ts and layout.ts, commit 99c007673d38f02c418d8e4cd2e9ec90b5d40be5. Original MIT notice: licenses/word-atlas-MIT.txt. WordMotion adds lifecycle/pixel budgets, filtering, accessible controls, localization integration and sharing.

WordMotion v18 self-hosted reading expansion (2026-10-31)
Six original-English prose editions from pinned official Standard Ebooks repositories: L. Frank Baum, The Wonderful Wizard of Oz; Frances Hodgson Burnett, The Secret Garden; Charles Dickens, A Christmas Carol; Jane Austen, Pride and Prejudice; Arthur Conan Doyle, The Adventures of Sherlock Holmes; Jack London, The Call of the Wild. Added 136 chapters; 8 books / 165 chapters total. Each book retains its Standard Ebooks LICENSE.md, source URL, source revision and a downloadable plain-text prose extraction with the notice. Edition contributors dedicate their work via CC0 1.0; original works are described by the source as believed public domain in the USA, without a blanket worldwide copyright determination. Front/back matter, third-party images and modern translations are not imported. Chinese introductions are original WordMotion AI-assisted learning material, not complete translations. Difficulty labels are suggestions, not certified CEFR levels. Device TTS remains separate from human recorded narration. No Cambridge textbook PDFs, mirrored commercial magazines or OpenStax NC material are imported.

WordMotion v19 self-hosted reading collection (2026-10-31)
100 additional original-English prose editions from pinned official Standard Ebooks repositories; total 108 books / 3,922 chapters. Per-edition source repository, revision, archive SHA256, LICENSE.md and downloadable prose extraction retain the Standard Ebooks CC0 contribution notice and its USA public-domain qualification. Author-name/death-year evidence for selected new editions was checked against the official Project Gutenberg offline CSV catalog (snapshot SHA256 3cece81ad3a9261f1b2fbfd2db1e1819cf0c4d3bcffc02ef74ccb8865b46895c). No modern translations, external artwork, or recorded narration are imported. New original books are labeled ungraded; Chinese display names are project translations, and short introductions are generated by WordMotion from author, genre and verified chapter/word counts. The books are historical works and retain their original language. Sources that failed author evidence, bounded archive size, or prose parsing checks were excluded from this batch. The reading shelf uses a compact local catalog, 20-book pages, English/Chinese search, genre/difficulty filters and per-book chapter manifests; individual chapters still load on demand.

WordMotion reading enrichment (2026-10-31): original English text remains unchanged. Chinese text is separately generated machine reference translation, not a published human translation. English narration is AI synthesized with Kokoro-82M (Apache-2.0) / kokoro-onnx (MIT), voice af_heart. Translation adapters support Helsinki-NLP/opus-mt-en-zh (Apache-2.0) and facebook/m2m100_418M (MIT); selected model, fixed revision and content checksums are retained in source/reading-enrichment/ and private server receipts. Incremental chapter JSON and MP3 files are independently hosted by the website, and actual per-book completion is displayed.
https://huggingface.co/hexgrad/Kokoro-82M
https://github.com/thewh1teagle/kokoro-onnx
https://huggingface.co/facebook/m2m100_418M
https://huggingface.co/Helsinki-NLP/opus-mt-en-zh

The production batch uses locally hosted Qwen2.5-7B-Instruct Q4_K_M (Apache-2.0), official GGUF revision bb5d59e06d9551d752d08b292a50eb208b07ab1f, via llama.cpp b11960 (MIT). The Alice chapter-one project AI translation sample is separately identified and retained as an immutable paragraph-hashed project asset. It is not labelled as output from the local Qwen model or human proofreading.
https://huggingface.co/Qwen/Qwen2.5-7B-Instruct-GGUF
https://github.com/ggml-org/llama.cpp