Speech Data Processor — NVIDIA NeMo SDP

nvidia.github.io

Tin mới

How to write config files?

How to add a new processor?

Supported datasets

Mozilla Common Voice Arabic (MCV)

MLS Italian (with P&C)

MLS Italian (no P&C)

MLS Spanish (no P&C)

Text MCV (Armenian)

Audio books (Armenian)

Librispeech (mini)

Kazakh Speech Dataset (KSD)

Kazakh Speech Corpus (KSC)

Kazakh Speech Corpus 2

Massive Arabic Speech Corpus (MASC)

Massive Arabic Speech Corpus (MASC): Extracting clean data from noisy train subset.

Tarteel AI’s Everyayah

Start Toloka for Armenian

Validation of responses Armenian

Getting final resuts from Toloka

TTS data processing pipeline

HiFiTTS-2 Bandwidth Estimation

Unlabeled Data Processing Pipeline

Earnings21/22 Dataset Processing Pipeline

Granary Dataset Creation Pipeline

Download source file