Uzbek/Russian Audio Datasets - Speech Restorised
Collection
Speech-restorised audio variants, including restored 48 kHz chunks. • 15 items • Updated
Datasets:
Request access to espeech_podcasts_chunked_speech_restorised
Access is gated because this repository contains speech data. Request access only for approved training, evaluation, or research use.
By requesting access, you agree not to redistribute files, attempt speaker re-identification, or use the data for unauthorized voice impersonation.
Log in or Sign Up to review the conditions and access this dataset content.
espeech_podcasts_chunked_speech_restorised
This is a gated Russian speech-restorised chunked speech dataset from
instinct-org. It contains byte-backed speech audio and transcripts for
speech-to-text training, evaluation, forced alignment, and data preparation
workflows.
STT Alignment Source
Use this repository as the current repaired source candidate for Russian podcast STT alignment. The matching base repository preserves the original transcript and segment metadata, but many later shards do not currently embed audio bytes. This speech-restorised repository keeps usable embedded audio bytes and can be consumed directly by forced-alignment and STT training pipelines.
Language
- Primary language:
ru(Russian)
Intended Use
- Russian speech-to-text training and evaluation after access approval
- forced alignment to generate token, word, character, and segment-level timing
- internal dataset curation, quality checks, and model evaluation
Data Notes
- Audio was processed through a custom internal speech restoration pipeline.
- This repository overlaps the source content of
espeech_podcasts_chunked; do not treat both repositories as independent samples in the same training mix. - Segment-level timing fields are source metadata. For training with aligned timestamps, regenerate alignments from this byte-backed audio and the transcript text.
- Some rows may have been excluded during restoration or quality filtering, so row counts are not expected to exactly match the base repository.
Access
This repository is gated with manual approval. Access requests are reviewed by the dataset maintainers. Users requesting access must provide their affiliation, intended use, and agreement to the dataset terms.
Collections
Safety And Rights
- Do not redistribute the files outside approved access.
- Do not attempt speaker re-identification.
- Do not use the data for unauthorized voice impersonation.
- Follow the source licenses and speaker-rights restrictions.
- Downloads last month
- 534