IndicVoices-R-Malayalam: 79.7 Hours of Malayalam Speech for TTS

IndicVoices-R-Malayalam is a subset of the IndicVoices-R multilingual text-to-speech corpus containing only Malayalam language data. The dataset comprises 31,106 audio samples totaling 79.74 hours of speech. It was created by the author 'trysem' and is hosted on Hugging Face, with a metadata timestamp of June 2026.

Use Cases

• Train a monolingual Malayalam text-to-speech model based on the 79.74 hours of audio data.
• Fine-tune a multilingual TTS model using the described Malayalam language subset.
• Benchmark speech synthesis quality for Indic languages using this specific corpus.
• Conduct linguistic analysis of Malayalam phonetics and prosody using the audio samples.

Strengths

• Contains 31,106 processed audio samples.
• Provides a substantial 79.74 hours of total speech duration.
• Derived from the described 'largest multilingual Indian TTS dataset', IndicVoices-R.

Limitations

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

Provenance

Source

Derived from the IndicVoices-R (IV-R) corpus, an automatic speech-derived multilingual Indian TTS dataset.

Collection Method

Likely contains audio recordings and corresponding text transcriptions for speech synthesis, derived via an automatic process.

Freshness

Last updated 2026-06-04 10:50:35; freshness should be verified.

Geography

Focus is on the Malayalam language, which is primarily spoken in the Indian state of Kerala.