Russian Podcast Audio and Transcripts Generated by NotebookLM

Russian-language podcast recordings generated by Google NotebookLM. The dataset contains 417 episodes totaling 77 hours, 23 minutes, and 22 seconds of audio in WAV format at 24 kHz. It was uploaded by AigizK to Hugging Face and last updated on March 26, 2026.

Use Cases

• Training text-to-speech models based on the male and female voice recordings.
• Fine-tuning automatic speech recognition models for Russian based on the provided audio and transcription pairs.
• Analyzing speech synthesis quality and prosody based on the generated podcast content.
• Creating synthetic training data for downstream NLP tasks based on the Russian transcriptions.

Strengths

• Contains 417 distinct audio episodes, providing a substantial number of samples.
• Offers a total duration of over 77 hours of Russian speech data.
• Includes both audio and full text transcriptions for each episode.
• Audio is provided in a high-fidelity WAV format at 24 kHz.

Limitations

• Column-level documentation is absent; field semantics for 'segments' must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale training.
• Data may reflect the inherent biases of the Google NotebookLM generation model and source material.

Provenance

Source

Generated by Google NotebookLM.

Collection Method

Audio podcasts were synthetically generated by the NotebookLM system.

Time Range

null

Freshness

Last updated 2026-03-26 07:53:03; freshness should be verified.

Geography

null