voices
/

VCTK_British_English_Males

Inference Endpoints

Model card Files Files and versions Community

VCTK_British_English_Males / prepare_model.py

jvision's picture

Updating speaker labels

b0e487c about 1 year ago

raw history blame contribute delete

2.98 kB

	import json
	import os
	import subprocess

	# Load the data from the provided dictionary
	data = {
	"VCTK_p226": {"age": 22, "gender": "M", "accents": "English", "region": "Surrey"},
	"VCTK_p227": {"age": 38, "gender": "M", "accents": "English", "region": "Cumbria"},
	"VCTK_p232": {"age": 23, "gender": "M", "accents": "English", "region": "Southern England"},
	"VCTK_p243": {"age": 22, "gender": "M", "accents": "English", "region": "London"},
	"VCTK_p254": {"age": 21, "gender": "M", "accents": "English", "region": "Surrey"},
	"VCTK_p256": {"age": 24, "gender": "M", "accents": "English", "region": "Birmingham"},
	"VCTK_p258": {"age": 22, "gender": "M", "accents": "English", "region": "Southern England"},
	"VCTK_p259": {"age": 23, "gender": "M", "accents": "English", "region": "Nottingham"},
	"VCTK_p270": {"age": 21, "gender": "M", "accents": "English", "region": "Yorkshire"},
	"VCTK_p273": {"age": 23, "gender": "M", "accents": "English", "region": "Suffolk"},
	"VCTK_p274": {"age": 22, "gender": "M", "accents": "English", "region": "Essex"},
	"VCTK_p278": {"age": 22, "gender": "M", "accents": "English", "region": "Cheshire"},
	"VCTK_p279": {"age": 23, "gender": "M", "accents": "English", "region": "Leicester"},
	"VCTK_p286": {"age": 23, "gender": "M", "accents": "English", "region": "Newcastle"},
	"VCTK_p287": {"age": 23, "gender": "M", "accents": "English", "region": "York"}
	}


	# Convert the data to JSON format
	json_data = json.dumps(data, indent=2)

	# Save the JSON data to a file
	with open('speakers-log.json', 'w') as file:
	file.write(json_data)

	# Run the TTS command to get the speaker indices
	command = "tts --model_path checkpoint_85000.pth --config_path config.json --list_speaker_idxs \| grep -vE '^(\s\\|\|\s>\|\s*$)'"
	output = subprocess.check_output(command, shell=True, text=True)

	# Parse the JSON output into a Python dictionary
	speaker_indices = eval(output)

	# Load the speaker IDs from speakers.json
	with open('speakers-log.json', 'r') as file:
	speaker_ids = json.load(file)


	for speaker_idx in speaker_indices:
	# # Remove the 'VCTK_' prefix
	speaker_id = speaker_idx
	# speaker_id = speaker_idx.replace('VCTK_', '')

	# Lookup the speaker ID in the loaded speaker IDs
	if speaker_id in speaker_ids:
	speaker_id_json = speaker_ids[speaker_id]
	else:
	continue

	# # Generate the TTS command to create the audio file
	# text = f"Hello, I am from {speaker_id_json['region']}. I hope that you will select my voice for your project. Thank you."
	# # make samples directory if it doesn't exist
	# if not os.path.exists("samples"):
	# os.makedirs("samples")

	# out_path = f"samples/{speaker_id}.wav"
	# tts_command = f"tts --text \"{text}\" --model_path checkpoint_85000.pth --language_idx en --config_path config.json --speaker_idx \"VCTK_{speaker_id}\" --out_path {out_path}"

	# Execute the TTS command
	# os.system(tts_command)