innovationm-ai/dailytalk-female

Dataset Summary

This dataset is a gender-specific subset of the original DailyTalk TTS dataset.
It contains English conversational speech paired with text transcripts, filtered and separated by speaker gender.

This version includes:

• Two columns:
  • text: transcription
  • audio: 24 kHz speech waveform
• One split (train) with 11,906 samples

No audio processing or text modifications were performed. The dataset is a structured subset of the original source.

Dataset Details

Uses

Direct Use

• Training or fine-tuning TTS models
• Building gender-specific voice models
• Emotion/style modeling
• Speaker-dependent research
• Voice cloning experiments that require consistent gender audio

Out-of-Scope Use

• Speaker identification or demographic inference beyond gender metadata
• Applications requiring balanced or diverse speaker populations
• Any high-stakes or biometric identification tasks

Dataset Structure

Data Fields

Field	Type	Description
text	string	Transcript of the spoken audio
audio	Audio(24 kHz)	Raw speech waveform

Split Information

The dataset contains a single train split.

Dataset Creation

Curation Rationale

This dataset was created to provide clean, gender-separated TTS data for training and evaluating models where consistent vocal characteristics are beneficial.

Source Data

The audio and text originate from the DailyTalk TTS dataset, which contains conversational English speech from multiple speakers.

Processing

• Filtered by speaker gender metadata from the original dataset
• No modifications to sampling rate, audio segments, or transcripts

Source Data Producers

Speakers from the DailyTalk dataset creators.
No demographic information beyond gender was used.

Bias, Risks, and Limitations

• Gender metadata comes from the original dataset; errors may propagate.
• The dataset reflects the biases, accents, and speaking styles of the original speakers.
• Not suitable for demographic or sensitive inference tasks.

Citation

If you use this dataset, please cite the original:

@dataset{innovationm2025dailytalk,
  title={DailyTalk TTS},
  author={InnovationM},
  year={2025},
  note={Hugging Face dataset}
}