/asm-speech

Creation of assamese speech dataset

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

asm-speech

Contains scripts for creating English and Assamese speech dataset.

Installation

Install Espeak-ng and FFmpeg

Use Python3.6+

pip install -r requirements.txt

Running

Create Syncmaps

python -m aeneas.tools.execute_task <abs_path_to_audio_file> <abs_path_to_text_file> "tts=espeak-ng|os_task_file_format=json|is_text_type=plain|tts_path=/usr/bin/espeak-ng|task_language=asm|allow_unlisted_languages=True" syncmap.json -r="allow_unlisted_languages=True|tts=espeak-ng|tts_path=/usr/bin/espeak-ng" --verbose

Audio chunking

python audio_chunk.py --audio-file <absolute_path_to_audio_file> --syncmap-file <absolute_path_to_syncmap_file>

Using finetuneas interface for fine tuning sync maps output.