capcode

Lossless normalization of text for the purpose of tokenization.

Fully UTF-8 Compliant

• Supports Unicode 13.0.0 (newer version supported depending on the implementation)
• Works correctly with any UTF-8 encoding scheme (NFC, NFD, etc.)

Features

• No information is lost
• The encoded text can be decoded exactly back to the original
• Safe: an LLM trained on this will still understand about uppercasing