Indian English and Hindi Speech Dataset for Text-to-Speech

73 minutes of curated, single-speaker speech data includes human-checked transcripts and quality control metrics. The dataset is hosted on Kaggle and appears designed for speech synthesis tasks. Specifics on the author, organization, and license are not provided.

Use Cases

• Train text-to-speech models based on the single-speaker audio recordings.
• Fine-tune speech synthesis systems for Indian English accents based on the described speech data.
• Develop Hindi speech synthesis systems based on the described audio content.
• Benchmark audio quality and transcription accuracy based on the provided QC metrics.

Strengths

• 73 minutes of curated speech data provides a substantial audio corpus.
• Human-checked transcripts likely ensure high accuracy for alignment tasks.
• Quality control metrics offer a basis for assessing data reliability.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset contains only a single speaker, limiting diversity for some model training scenarios.

Provenance

Source

kaggle

Collection Method

Curated speech collection, method details unspecified.

Geography

Likely India, inferred from language focus.