Pygmalion to TextGenWebUI JSON Conversion

The migration from **Pygmalion** to **TextGenWebUI** (oobabooga) represents a technical shift from foundational, prefix-heavy JSON standards to a modern, context-driven inference platform. While Pygmalion served as the classic bridge for numerous early roleplay tools, TextGenWebUI requires character logic to be consolidated into its specific context and greeting schema to drive model behavior efficiently across its diverse model loader ecosystem.

1. Deconstructing Legacy Schema Prefixes

The Pygmalion schema is known for its foundational key prefixes like char_persona and world_scenario. TextGenWebUI thrives when this character logic is synthesized into a cohesive context monolith. Our converter identifies these classic prefixes and performs a "Semantic Synthesis," merging the character's behavioral goals and world-building cues into the standardized context block that oobabooga uses to set the initial model state, ensuring your characters remain behavioral consistent during the transition.

2. Optimizing for Modern Local Inference

Legacy Pygmalion characters were often authored for smaller models with limited context window sensitivities. TextGenWebUI is designed for the absolute depth of model orchestration, often running state-of-the-art LLMs (Llama, Mistral, Command R). Our conversion engine identifies the most high-impact traits from your Pygmalion source and priorities them in the resulting JSON, ensuring that your legacy digital personas remain detailed and coherent without overwhelming the local LLM's memory, resulting in faster inference and more relevant narrative responses.

3. Dialogue and Greeting Normalization

We map the Pygmalion char_greeting directly to the TextGenWebUI greeting field. Furthermore, we reformat the example_dialogue block into standardized patterns that maximize its influence on the character's speaking style. This is critical because TextGenWebUI supports a vast array of model formats (GGUF, EXL2) that interpret dialogue cues differently. By standardizing the output, we ensure a consistent speaking style and personality regardless of the specific model you choose to load into the webui.

4. Privacy and the "Zero-Server" Advantage

Character development represents a significant creative investment. We honor this by performing all extractions and mapping locally. By running the conversion logic entirely in your browser's RAM, we ensure that your proprietary character definitions are never transmitted to our server or any third party. This provides a secure, private bridge for users who want to transition their legacy digital assets to a more modern, local-first inference environment.

5. Identity Integrity and Metadata Generation

When you export to TextGenWebUI, our tool generates a standardized JSON file that is ready for instant use in your characters folder. This ensures that your Pygmalion digital assets are integrated perfectly with the webui's library management features, facilitating easier organization and allowing you to leverage the full power of local model orchestration without losing the nuanced personalities you've crafted for your legacy roleplay experiences.