Credits

AI & Language Processing

• Google Gemini — Powers AI translations, word-by-word breakdowns, contextual explanations, and image OCR (text extraction from screenshots).
• Claude Code by Anthropic — Used for dictionary content generation, including definitions, example sentences, character breakdowns, tips, and quality assurance. Also assisted in building the web dictionary and the mobile app.
• CosyVoice 3 — Open-source text-to-speech model from Alibaba's FunAudioLLM team. We self-host it to generate example sentence audio with a Beijing Mandarin voice. Apache 2.0. github.com/FunAudioLLM/CosyVoice
• Microsoft Azure Speech — Neural text-to-speech used for single-word pronunciation audio, with forced pinyin so characters with multiple readings are pronounced correctly.
• WhisperX — Used for character-level alignment between generated audio and source text, powering the karaoke-style highlighting on word and example pages. github.com/m-bain/whisperX

Dictionary Data

• SUBTLEX-CH — Word frequency data based on Chinese film and television subtitles (33.5 million words). Used for frequency rankings in the dictionary.
  Cai, Q., & Brysbaert, M. (2010). SUBTLEX-CH: Chinese Word and Character Frequencies Based on Film Subtitles. PLOS ONE, 5(6), e10729.
• HSK 3.0 Exam Syllabus — We tag dictionary entries with their HSK level from the official word list in 《中文水平考试 HSK 考试大纲》, published November 2025 by the Center for Language Education and Cooperation (中外语言交流合作中心) and in effect from July 2026. Published at chinesetest.cn.
• HanziWriter — JavaScript library for Chinese character stroke order animations. Published under the MIT License. hanziwriter.org
• Make Me a Hanzi — Character stroke order data (the animations and writing practice in the app and on word pages), plus the character decomposition and radical data behind the components breakdown. Stroke data extracted from fonts by Arphic Technology. Published under the Arphic Public License. GitHub

References we check against

The p8nda dictionary is written from scratch. These are the references our generation and review process checks entries against, mostly for pinyin and traditional forms. None of them is the source of the dictionary.

• CC-CEDICT — Community-maintained Chinese-English dictionary. Licensed under CC BY-SA 4.0.
• chinese-xinhua — Open dataset of Xinhua Dictionary (新华字典) entries, also used for per-character reference data. MIT.
• THUOCL — Open Chinese lexicon from Tsinghua University's NLP lab, used to validate chengyu coverage and pinyin. MIT.

Technology

• React Native & Expo — Mobile app framework. Built with Expo SDK 55 and expo-router.
• ASP.NET Core — Backend API and web dictionary. Handles translations, OCR, authentication, and data sync.
• Hangfire — Background job processing for AI translations and OCR requests.
• PostgreSQL — Server-side database for user accounts, translations, OCR jobs, and sync data.
• SQLite (expo-sqlite) — On-device database powering offline dictionary search and local storage.
• Cloudflare — DNS, CDN, and secure tunneling for the production API.
• Umami — Privacy-focused, cookie-free, open-source web analytics. Self-hosted. umami.is
• Altcha — Open-source bot protection on the sign-in, registration, and password reset forms. Your browser solves a small puzzle in the background, so there are no image grids to label and no tracking by a third party. altcha.org

Design & Icons

• Ionicons — Icon set used throughout the app and website. ionic.io/ionicons
• Tailwind CSS — Utility-first CSS framework used for the website and admin dashboard.