Localization is a permanent part of your product. Quetzal is the layer that keeps it alive, so engineers don't have to think about it.
Quetzal isn't built for a specific stack, or a specific framework. Your engineers still choose the mechanisms that work best, but the labor of maintaining it falls completely on us.
It connects to the workflow you already have and handles the parts that tend to slip: translating new strings, checking existing translations against real context, and finding user-facing text that was never wired up for localization in the first place.
This lets you localize with the confidence that the maintenance won't create an endless burden for your engineers.
The most difficult part of localization was never the execution of the translation itself. It was always in providing the translator with proper context.
With Quetzal, the rich context from your code is distilled to ensure accurate, consistent translation. Translations are performed or reviewed considering similar text, surrounding context, brand voice, glossary terms, and locale specific rules.
This effect compounds the more context Quetzal has. If you link your code with your documentation, Quetzal will cross-reference your code to ensure consistency across the board.
Localization is closer to an art than a science. A team that knows your product will always do it better than a model, and we don't expect that to change. Human language will always be for humans.
This is why Quetzal isn't just a translation service. It's built to find the nuance that human translators and engineers miss, so localization quality improves regardless of the shape of your team. It won't fight with humans in the loop, it will just pick up their slack.
The vigilance and attention to detail even across huge datasets makes AI an essential partner for localization efforts. While AI translation is a strong starting place, then endgame is AI augmented translation, and Quetzal hopes to play that role for your team.
In the past couple decades, the software industry has settled on the pattern of using an external service to translate on a string-by-string basis. This pattern of treating translators like machines, and translation as a word-by-word affair has made the majority of software translation riddled with clunky wording and consistency issues.
Users are used to this, and it makes it difficult to quantify the cost of poor translation. They will often struggle silently to navigate, only churning if a particular setting or error becomes too tricky to parse. In this way, the cost stays invisible, and an invisible cost is one no one is paid to fix.
Our goal is to raise the standard. With the rich context gathering capabilities of AI, there is no excuse for the clunky mistakes we have grown used to.
Link your repo once. Quetzal keeps it localized from there.