A RIME plum recipe and configuration for typing Quick5 (速成) efficiently.
This configuration enables continuous Quick5 typing (“速成連打”), allowing you to type words and phrases fluidly without selecting individual characters after each code. It supports both English (US) and German (DE) keyboard layouts and extends Quick5 typing to cover Japanese Kokuji, Korean Gukja, and Vietnamese Chữ Nôm characters.
Recipe: ℞ easonwong-de/rime-eason-quick5
Install this recipe and its dependencies using plum (rime-install):
bash rime-install cangjie essay Iorest/rime-dict easonwong-de/rime-eason-quick5:installThis installs all required upstream dependencies alongside this recipe:
- rime-cangjie (Cangjie 5 dictionary)
- rime-essay (Standard Chinese phrase language model)
- rime-dict (Extended Chinese phrase dictionary)
After installation, launch RIME’s settings menu (or trigger your frontend's deployment) to apply the configuration.
-
Continuous Typing (“速成連打”) Unlike traditional Quick5 setups where you must confirm each character individually, this configuration supports uninterrupted typing—words and phrases are automatically composed as you type.
-
Multi-Language Keyboard Support Supports English (US) and German layouts (affecting punctuation only, not the Quick5 arrangement). Switch via
F4: select英鍵速成for US or德鍵速成for German keyboards.
-
Extended Character Support Most Chinese characters are supported, as well as many Japanese Kokuji, Korean Gukja, and Vietnamese Chữ Nôm can be typed out using Quick5.
-
Additional Scripts Japanese Kana, Korean Hangul, Bopomofo, and Greek letters can be typed using the symbol hotkey
符.
-
Emoji Suggestions Single-letter Cangjie radical emojis work out of the box. To enable full keyword suggestions:
bash rime-install emoji bash rime-install emoji:customize:schema=quick5 bash rime-install emoji:customize:schema=quick5de
-
Character Conversion Easily toggle conversion to Simplified Chinese or Japanese Shinjitai via the
F4menu.
- Cantonese Vocabulary (
cantonese.dict.yaml): Sourced and adapted from CanCLID/rime-cantonese-upstream by the Cantonese Computational Linguistics Infrastructure Development Workgroup (CanCLID), licenced under CC BY 4.0. Vocabulary entries are extracted and filtered from upstream datasets (word.csv,fixed_expressions.csv,phrase_fragment.csv, andtrending.csv).