Emilia S2S Mimi Q8 Named Speakers: A TinyAya Codec-Tokenized TTS Dataset

A text-to-speech dataset built from TinyAya codec-tokenized examples. The dataset contains named speaker examples for Ira, Aisha, Siya, and Zoya, with training text that includes speaker prefixes. It was created by rumik-ai and last updated on 2026-05-26.

Use Cases

• Train or fine-tune TTS models based on the codec-tokenized audio format described.
• Develop multi-speaker TTS systems based on the named speaker examples for Ira, Aisha, Siya, and Zoya.
• Benchmark speech synthesis quality using the provided speaker-specific audio-text pairs.

Strengths

• Includes explicit speaker mapping for four distinct voices: Ira, Aisha, Siya, and Zoya.
• Training text fields are pre-formatted with speaker prefixes, which may simplify model input preparation.

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.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Collection Method

Built from gs://gp-s2s-data/s2s-data/emilia-s2s/chunk_00036_200k/parquet.

Freshness

Last updated 2026-05-26 13:37:09; freshness should be verified.