528.2 hours of filtered Russian speech data across the audiobook genre. The corpus is processed through the BALALAIKA pipeline by the MTUCI lab260 team for generative speech tasks.
Use Cases
- Train text-to-speech (TTS) models using the Russian audio recordings and their corresponding text annotations
- Develop speech synthesis systems optimized for the audiobook genre using the 528.2 hours of filtered speech
- Fine-tune generative speech models using the high-quality Russian speech corpus
Strengths
- 528.2 hours of total audio duration after filtering from an initial 1000+ hours
- Annotated using the BALALAIKA pipeline developed by the MTUCI lab260 team
- Sourced exclusively from the Russian audiobook genre
- Released under the apache-2.0 license