Jenny 30H Tokens: Pre-tokenized Audio Codec Tokens for TTS Training

somu9 provides 20,141 pre-extracted audio codec tokens for text-to-speech training, derived from the reach-vb/jenny_tts_dataset. The collection contains 26.4 hours of audio, tokenized using the MOSS-Audio-Tokenizer-Nano codec at 48 kHz stereo and a frame rate of 12.5 Hz.

Use Cases

• Fine-tuning TTS models based on pre-extracted audio tokens.
• Training audio codec language models using the provided token sequences.
• Benchmarking tokenizer performance on the 26.4-hour audio corpus.
• Synthesizing speech from the tokenized representations of the source dataset.

Strengths

• Contains 20,141 individual audio samples.
• Totals 26.4 hours of audio data.
• Uses a specific audio codec (MOSS-Audio-Tokenizer-Nano) at 48 kHz stereo.
• Provides an average sample duration of 4.7 seconds.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.
• The source dataset's geographic and demographic biases are unknown.

Provenance

Source

Derived from reach-vb/jenny_tts_dataset.

Collection Method

Pre-extracted using MOSS-Audio-Tokenizer-Nano codec.

Freshness

Last updated 2026-05-18 05:32:34; freshness should be verified.