195 AI character personas have voice designs, generated dialogue, and synthesized speech based on their character settings. The dataset contains 20,800 audio files, including 195 reference voices and 20,600 spoken utterances, all in WAV format. It was created by kizuna-intelligence and last updated on April 11, 2026.
Use Cases
- Train Japanese text-to-speech models based on the 195 distinct character voice profiles.
- Develop voice cloning systems for AI personas using the reference and spoken utterance audio files.
- Generate emotional or descriptive speech for virtual characters based on the 'original', 'descriptive', and 'emotional' dialogue categories.
- Benchmark audio synthesis quality across a large set of unique, character-driven Japanese voices.
Strengths
- Contains 20,800 high-fidelity audio files in WAV format with PCM 16-bit, 44.1kHz, mono specifications.
- Covers 195 distinct AI character personas, providing a broad range of voice profiles.
- Dialogue is categorized into 'original', 'descriptive', and 'emotional' types, suggesting structured content.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count, file formats beyond WAV, and license information are unknown, which may limit suitability assessment.
Provenance
- Source
- kizuna-intelligence via Hugging Face.
- Collection Method
- Voice design, dialogue generation, and audio synthesis performed for each AI character persona based on their settings.
- Time Range
- null
- Freshness
- Last updated 2026-04-11 08:41:46; freshness should be verified.
- Geography
- null