IITM Mono Hindi Female: Single-Speaker Hindi Speech Synthesis Dataset

A monolingual Hindi text-to-speech dataset containing 6,926 utterances from a single female speaker. The audio data is embedded in parquet files at a 48kHz sampling rate and was extracted from the IndicTTS project by SPRINGLab at IIT Madras. The dataset was uploaded to Hugging Face by the user 'somu9'.

Use Cases

• Train a text-to-speech model for Hindi based on the single female speaker's voice.
• Fine-tune a pre-trained TTS model using the 48kHz mono audio and text pairs.
• Benchmark speech synthesis quality for Hindi using the structured parquet shards.
• Develop voice cloning applications for Hindi, leveraging the consistent single-speaker audio.

Strengths

• Contains 6,926 audio-text utterance pairs.
• Audio is high-quality 48kHz mono, embedded in a structured parquet format.
• Originates from a known academic project (SPRINGLab/IndicTTS-Hindi at IIT Madras).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset description is minimal; actual data quality requires manual inspection after download.

Provenance

Source

SPRINGLab/IndicTTS-Hindi project at IIT Madras.

Collection Method

Extracted from the IndicTTS project; specific gathering method is not detailed.

Freshness

Last updated 2026-04-15 11:38:55; freshness should be verified.