Fizzo Novel Language Change

Working with Fizzo Novel Language Change for Manuscript Rewrites

I spent last year helping a small indie publisher migrate three completed manuscripts through a language conversion pipeline using Fizzo Novel Language Change. The project seemed straightforward on paper — shift from American English to British English across 180,000 words each. What I learned in those six months would have saved us about eighty hours if someone had told me upfront. The tool works by first ingesting your manuscript file, parsing it into individual chapters or sections, applying a configurable language rule set, and then outputting a revised document. The initial setup involves installing the CLI package, running the configuration wizard, and defining your source and target language variants. The default presets cover major regional English dialects and several European languages, but the real power comes from writing custom rule sets. Rule sets are YAML files that define word-level substitutions, grammar adjustments, punctuation shifts, and style overrides. A basic American-to-British set includes color-to-colour, realize-to-realise, and similar orthographic swaps. Where most people stop is where the actual work begins. The tool also needs rule files for regional idioms, metric conversions, and register shifts — none of which are included out of the box.

The Actual Conversion Workflow

Here is how the process runs in practice. You place your manuscript files in an input directory, write your rule set, and execute the conversion command. The tool processes chunks sequentially, so a 90,000-word novel typically takes between four and twelve minutes depending on your machine and rule complexity. Output lands in a separate directory with sidecar metadata files that log every substitution made. The metadata logging is honestly the most useful part. When your editor comes back with forty questions about whether "apartment" was correctly converted in chapter seven, you open the sidecar file and find the exact line. Without it, you would be doing manual diff sweeps through the entire manuscript, which is what we did on the first book. Took us two days.

Edge Cases That Will Waste Your Time

Let me tell you about the one that nearly cost us a week. A romance novelist in our batch had written extensive dialogue with phonetic spellings — characters pronounced words in ways that deliberately conflicted with standard spelling. When Fizzo processed the language change rules, the phonetic variants got "corrected" to match the target dialect's standard forms. The character who was supposed to sound working-class suddenly sounded like they'd gone to boarding school. Every phonetic deviation in dialogue was silently normalized. The workaround was to tag dialogue sections as protected using XML annotations around speech passages, then configure the rule engine to skip annotated blocks. We wrote a pre-processing script that wrapped all dialogue in `` tags, ran the conversion, then stripped the tags from the output. It added about three minutes to the pipeline per manuscript, but it was faster than manually reviewing every chapter for dialect drift.

Counter-Intuitive Things I Wish I Knew Earlier

First, more rules do not mean better output. We started with a rule set containing over 3,000 entries and found that the quality actually dropped. The engine began applying rules out of intended order, creating cascading conflicts where one substitution triggered another unintended one. We trimmed the set down to 847 carefully ordered entries and the output became significantly cleaner. Rule ordering matters more than rule count, and duplicate entries across categories cause silent regressions. Second, the tool handles proper nouns inconsistently by default. Character names, place names, and brand references get caught in blanket substitution rules unless you explicitly whitelist them. On the second manuscript, an author named their protagonist "Ashford" and the rule set converted it to "Ashfurd" in British mode because of an archaic spelling variant in the dictionary file. We had to build a personal glossary file that overrides all rule-set behavior for listed terms. This is not well-documented in the default help output.

Performance Reality Check

Fizzo Novel Language Change is fast for straightforward regional dialect shifts. American to British English, Canadian to Australian English — these run clean with minimal manual review. Where it struggles is literary translation adjacent work. Shifting register, dialect, or voice within the same broad language family produces uneven results that require heavy human editing. We estimated that manuscripts intended for a full dialect shift — say, Southern American English to Cockney — needed roughly sixty percent human rewriting after the tool ran. The memory profile is also worth noting. Processing large files pushes RAM usage to about 2.4 gigabytes for a single-pass run on a 100,000-word manuscript. If you are running this on a machine with less than 8GB available, you should chunk the files beforehand or you will see slowdowns and occasional silent truncations at the end of long chapters.

Download and Installation

The tool is distributed through npm and PyPI. For the Node version, run `npm install -g fizzo-language-change`. The Python binding is `pip install fizzo-novel-lang`. Both require Node 18 or Python 3.11 minimum. Documentation lives at the official repository, which includes the rule set schema reference and the protected block syntax we used. There is no web interface — everything is CLI-driven, which is fine once you get comfortable with it. If you are just doing a quick regional swap on a short story, the built-in presets handle it in under two minutes. If you are running a full-length novel through a custom rule set with protected sections and a personal glossary, budget three to five hours of setup and review per manuscript. The tool itself does the heavy lifting, but the configuration work is non-trivial and poorly mapped in the default documentation.