Synthetic AI VTuber Live Chat Conversations in Japanese

1,946 synthetically generated chat sessions simulate a four-turn interaction between an AI VTuber and multiple viewers. The dataset was created in March 2026 by DataPilot using the SDG-Nexus pipeline, blending AI VTuber personas with user personas from the nvidia/Nemotron-Personas-Japan model.

Use Cases

• Train a Japanese dialogue model to generate AI VTuber responses based on simulated viewer comments.
• Analyze interaction patterns within the four-turn chat structure to model conversational flow.
• Fine-tune a persona-conditional language model using the blended AI VTuber and user persona inputs.
• Study synthetic chat log characteristics for human-computer interaction research in live-streaming contexts.

Strengths

• Contains 1,946 complete, multi-turn chat sessions.
• Data is synthetically generated with a defined pipeline (SDG-Nexus), ensuring structured format.
• Explicitly designed for a specific domain: AI VTuber and viewer interactions in Japanese.

Limitations

• All data is synthetic, not collected from real-world streams, limiting realism for some applications.
• The fixed four-turn structure may not reflect the variability of actual live chat durations.
• Lacks metadata on specific columns, features, or sample data for detailed inspection.

Provenance

Source

DataPilot on Hugging Face.

Collection Method

Synthetically generated using the SDG-Nexus pipeline, combining AI VTuber personas and nvidia/Nemotron-Personas-Japan.

Time Range

Synthetic generation date is March 2026.

Freshness

Generated and published in March 2026.

Geography

Language and cultural context is Japanese.