Good Tools — Text & language engines (26 October 2026)

All conversion runs on the visitor's device. No fonts are redistributed.

1. Bengali Bijoy / SutonnyMJ
@abdalgolabs/ansi-unicode-converter 1.0.5 (MIT).
https://www.npmjs.com/package/@abdalgolabs/ansi-unicode-converter
https://github.com/Abdalgolabs/ansi-unicode-converter
Source (CommonJS converted to a standalone ES module; conversion tables/algorithms retained):
https://usegoodtools.com/text-language/vendor/bijoy.mjs
Good Tools corrections: verified Sutonny conjuncts, dash mapping, transient joiner preservation.
Licence: https://usegoodtools.com/text-language/vendor/Bijoy-LICENSE.txt

2. Hindi Kruti Dev 010
Adapted from Deepak Kamboj indianlanguageconverter, commit
12a87e365b0c2ee6e513343c0b740ac6a8c7d6d6.
https://github.com/deepakkamboj/indianlanguageconverter
The upstream project declares conversion algorithms under GPL-2.0.
Good Tools distributes this standalone conversion engine under GPL-2.0.
Changes: bounded substitutions, full-cluster matra/reph reordering, NFC,
nukta reverse codes, typographic quote fixes for distinct sha/ssa, chai/chau.
The complete preferred source form is the served JavaScript module:
https://usegoodtools.com/text-language/vendor/krutidev.mjs
Licence: https://usegoodtools.com/text-language/vendor/KrutiDev-LICENSE.txt
Reference tables: SIL KrutiDev011.map (includes Kruti Dev 010 notes).
https://github.com/silnrsi/wsresources/blob/master/scripts/Deva/legacy/kruti-dev-011/mappings/KrutiDev011.map
Supplemental SIL International (2006) mapping data, MIT:
https://usegoodtools.com/text-language/vendor/SIL-LICENSE.txt
No font is included. DevLys and Chanakya are not supported.

3. Chinese
OpenCCJS 1.4.2 core (MIT); official OpenCC dictionary data (Apache-2.0).
Core source: https://usegoodtools.com/text-language/vendor/opencc-core.mjs
Runtime: https://usegoodtools.com/text-language/opencc-runtime.mjs
Data manifest and SHA-256 hashes:
https://usegoodtools.com/text-language/opencc-data-manifest.mjs
All dictionaries are pinned to official OpenCC revision:
a806b775d038f9fa7c061f50d270115e7ad7da5b
https://github.com/BYVoid/OpenCC/tree/a806b775d038f9fa7c061f50d270115e7ad7da5b/data/dictionary
The runtime builds tables from those files and follows OpenCCJS config pipelines,
including regional projections, reverse-variant preferences and extension characters.
Reference result: all 392 applicable official regression vectors pass.
OpenCCJS source: https://github.com/nk2028/opencc-js
MIT: https://usegoodtools.com/text-language/vendor/OpenCC-LICENSE.txt
Apache-2.0: https://usegoodtools.com/text-language/vendor/Apache-2.0.txt
Third-party notices: https://usegoodtools.com/text-language/vendor/OpenCC-NOTICES.txt
Fourteen pinned dictionary texts (about 1.3 MB total) load from jsDelivr with
SHA-256 checks, credentials omitted and no input text sent. First load needs
internet access; later conversions reuse the engine in the current tab.
Hong Kong uses character variants; experimental HK vocabulary is disabled.

4. Japanese
Good Tools selective Unicode-range conversion. Uses targeted NFKC only on
halfwidth kana runs, NFC only on fullwidth kana runs. Other scripts and
compatibility characters are not normalized.
https://unicode.org/reports/tr15/
Source: https://usegoodtools.com/text-language/core.mjs

General limits
Plain text carries no font metadata. Legacy ASCII cannot be distinguished
from ordinary Latin text reliably: the user must choose whole legacy input
or select a legacy passage. Unicode script runs can be identified and are
preserved or converted by direction. Encoding conversion is not translation,
OCR, font installation or preservation of rich document formatting.