Voiceconvdesign: Conversational AI Audio Dataset with TTS Prompts

Voiceconvdesign is a dataset of conversational turns generated for text-to-speech (TTS) voice design. It contains structured records linking system prompts, agent transcripts, and synthesized audio clips. The dataset was authored by ShiniChien and last updated on Hugging Face on 2026-05-12.

Use Cases

• Train or fine-tune TTS models based on the provided audio and transcript pairs.
• Analyze the relationship between system prompts and generated speech characteristics based on the 'prompt' and 'voice' columns.
• Study conversational flow and turn-taking dynamics in synthesized dialogues based on the 'turn_index' and 'conversation_id' structure.
• Benchmark audio generation quality and duration based on the 'duration_s' and 'audio' columns.

Strengths

• Includes raw audio data in a standard WAV format (24 kHz mono 16-bit).
• Provides the full system instruction ('prompt') used to generate each turn, offering insight into the generation process.
• Structures conversations with unique identifiers and turn indices, preserving dialogue context.

Limitations

• Dataset scale is unknown; row count, file size, and sample data are not provided.
• Column-level documentation is absent; field semantics beyond basic names must be inferred after download.
• Last updated 2026-05-12 05:12:27; freshness should be verified.

Provenance

Source

ShiniChien via Hugging Face.

Collection Method

Likely generated using the Gemini Live API for conversational TTS synthesis.

Freshness

Last updated 2026-05-12 05:12:27.