KoboldAI to Voxta Character Conversion

Converting a character from **KoboldAI** to **Voxta** is a process of transitioning from a backend-focused storage model to a frontend-focused companion model. KoboldAI's JSON format was designed primarily to save the state of a model's context, whereas Voxta's schema is designed to define a character's "Interactive Identity," including voice, visual profile, and behavioral constraints.

1. Remapping Key Schema Differences

KoboldAI JSON files (often used in KoboldCPP or legacy Lite frontends) typically use a flat structure with specific naming conventions for context. Common keys include description, scenario, and first_mes. Voxta, a more modern and specialized application, requires a specific JSON hierarchy that often uses PascalCase or CamelCase keys like Name, FirstMessage, and Personality.

Our conversion tool handles this transition by performing a "Semantic Scan" of the source JSON. We don't just rename keys; we identify the purpose of the text within those keys. For example, the description in Kobold often includes both the character's physical traits and their personality. We attempt to parse these out to populate both the profile and personality fields in Voxta, which allows the Voxta engine to better calibrate its multimodal responses.

2. Handling "World Info" and Lorebooks

One of the most complex aspects of KoboldAI characters is their reliance on "World Info" (WI). These are essentially keyword-triggered lore entries. Voxta handles character knowledge through a distinct "Memory Book" system. While our converter primarily targets the core persona, it ensures that the character's most vital world-building elements are preserved within the scenario or description blocks of the new Voxta JSON, ensuring that the character doesn't lose its "historical context" during the migration.

3. The Importance of UUID and Metadata

KoboldAI saves are often identified by their file name. Voxta, however, relies on a globally unique identifier (UUID) to manage its character database and prevent profile collisions. Our converter automatically generates a compliant UUID for your character during the process. Additionally, we populate the Voxta version and creatorNotes fields to ensure that your character is treated as a production-ready asset from the moment it is imported.

4. Dialogue Sanitization for TTS Engines

Because Voxta characters are designed for voice interaction, the way mes_example (Example Dialogue) is formatted is crucial. KoboldAI files often contain dialogue examples formatted for raw text-completion models, which can include messy tokens or inconsistent quotation styles. Our engine cleans these dialogue strings, removing artifacts that might cause "hiccups" in a Text-to-Speech (TTS) engine, while maintaining the character's unique cadence and speaking style.

5. Why Browser-Side Processing Matters

By executing this conversion locally using JavaScript, we provide a level of security that cloud-based converters cannot match. Your character's personality, secret lore, and custom instructions are processed entirely in your computer's RAM. No data is ever transmitted to our servers, ensuring that your creative work remains entirely your own. This "Zero-Server" approach is essential for creators who value privacy and data sovereignty in the age of AI.