clarvia-tts-tiny / config.json
ekremcet's picture
Add the PyTorch weights as model.safetensors, with the architecture config
867322d verified
Raw History Blame Contribute Delete
1.87 kB
{
"sample_rate": 24000,
"hop": 256,
"voices": [
"male_1",
"male_2",
"female_1",
"female_2",
"female_3",
"female_4",
"female_5",
"male_3",
"male_4",
"male_5",
"unused"
],
"symbols": [
"<pad>",
"<bos>",
"<eos>",
"<sp>",
"<unk>",
",",
".",
"?",
"!",
";",
":",
"\"",
"'",
"(",
")",
"—",
"…",
"p",
"b",
"t",
"d",
"k",
"ɡ",
"c",
"ɟ",
"f",
"v",
"s",
"z",
"ʃ",
"ʒ",
"h",
"m",
"n",
"ŋ",
"l",
"ɫ",
"ɾ",
"j",
"tʃ",
"dʒ",
"a",
"e",
"i",
"ɯ",
"o",
"ø",
"u",
"y",
"aː",
"eː",
"iː",
"oː",
"øː",
"uː",
"yː"
],
"max_tokens": 512,
"max_frames": 3000,
"step": 190000,
"frontend": "clarvia-phonemizer 0.1.0",
"inputs": [
"tokens",
"stress",
"voice",
"speed",
"pitch_shift",
"pitch_range"
],
"outputs": [
"audio",
"durations"
],
"public_voices": [
"male_1",
"male_2",
"female_1",
"female_2",
"female_3",
"female_4",
"female_5",
"male_3",
"male_4",
"male_5"
],
"model": {
"sample_rate": 24000,
"hop": 256,
"n_fft": 1024,
"n_mels": 80,
"f0_min": 50.0,
"f0_max": 600.0,
"speakers": [
"male_1",
"male_2",
"female_1",
"female_2",
"female_3",
"female_4",
"female_5",
"male_3",
"male_4",
"male_5",
"unused"
],
"vocab_size": 56,
"channels": 192,
"heads": 4,
"feedforward": 640,
"kernel": 7,
"encoder_layers": 4,
"prosody_layers": 2,
"acoustic_layers": 5,
"dropout": 0.1,
"predictor_channels": 128,
"synth_fft": 512,
"synth_hop": 128,
"noise_fft": 256,
"noise_hop": 64,
"cep_causal": 80,
"cep_anti": 16,
"noise_cep": 32,
"harmonic_limit": 0.47,
"max_tokens": 512,
"max_frames": 3000,
"mel_mean": -5.0,
"mel_scale": 5.0,
"energy_mean": -4.0,
"energy_scale": 2.0,
"pitch_reference_hz": 200.0
}
}