Game Development
Multilingual in-game text localization
Mixed Realms needed high-quality translation, because even minor errors can harm user experience, plus they needed a solution that could cover several languages at once. The team had been localizing their narrative-heavy action RPG, 13Z: The Zodiac Trials, into English, Chinese, and Japanese through vendors and spreadsheets, without in-house native speakers to catch coherence, tone, or placeholder issues. Off-the-shelf machine translation got them to roughly 80-90% usable Mandarin output, but struggled with character voice, narrative coherence, and technical tags. They wanted to expand into secondary languages - German, Russian, Spanish, French, and Portuguese - in a cost-efficient way, without sacrificing the quality that a narrative-driven game world demands.
Ran a hands-on proof of concept in the Algebras playground, comparing agentic translation against straight machine translation with EN/ZH prompting, so the team could judge real quality differences before committing to a plan.
Connected Algebras CLI directly to Mixed Realms' existing Google Sheets + Unity (I2 Localization) pipeline, turning source strings into ready-to-use locale files without building a custom API integration.
Protected in-game placeholders and tags (e.g. @key/text@) through custom prompt rules, and used Algebras CLI's multi-source-language config support (EN to ES/DE/RU/FR/PT, ZH to JA/KO) so the team could translate from multiple source languages without recreating their configuration each time.
Gave the team agentic (cultural-layer) translation and glossary support via API to keep tone and terminology consistent across narrative-heavy content, with responsive, hands-on support for fast troubleshooting along the way.
Localization quality rated 7-8/10 by native-speaking players, with the Spanish version specifically scoring 7.5/10 and no game-breaking mistakes reported.
The team translated their full ~25k-word text base into German on a tight turnaround, relying on Algebras' translation history and caching to avoid re-processing and unnecessary cost on reruns.
Multi-source-language configs, glossary access via CLI, and built-in client-side rate-limit handling let the team run localization batches across five languages without manual workarounds.
Before Algebras: Mixed Realms depended on external vendors and manual spreadsheet hand-offs for English, Chinese, and Japanese, with no in-house native speakers to catch tone, coherence, or placeholder issues in machine-translated drafts.
After Algebras: A CLI-driven workflow plugged directly into their existing Google Sheets/Unity pipeline, using prompt rules and glossaries to protect placeholders and tone - cost-effectively covering German, Russian, Spanish, French, and Portuguese alongside their in-house English, Chinese, and Japanese work.
.svg)
.avif)
