Text-to-Speech
Transformers
ONNX
Safetensors
English
Chinese
automatic-speech-recognition
voice-conversion
speech
audio
Instructions to use AutoArk-AI/GPA with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use AutoArk-AI/GPA with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-to-speech", model="AutoArk-AI/GPA")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("AutoArk-AI/GPA", dtype="auto") - Notebooks
- Google Colab
- Kaggle
| audio_tokenizer: | |
| mel_params: | |
| sample_rate: 16000 | |
| n_fft: 1024 | |
| win_length: 640 | |
| hop_length: 320 | |
| mel_fmin: 10 | |
| mel_fmax: null | |
| num_mels: 128 | |
| encoder: | |
| input_channels: 1024 | |
| vocos_dim: 384 | |
| vocos_intermediate_dim: 2048 | |
| vocos_num_layers: 12 | |
| out_channels: 1024 | |
| sample_ratios: [1,1] | |
| decoder: | |
| input_channel: 1024 | |
| channels: 1536 | |
| rates: [8, 5, 4, 2] | |
| kernel_sizes: [16,11,8,4] | |
| quantizer: | |
| input_dim: 1024 | |
| codebook_size: 8192 | |
| codebook_dim: 8 | |
| commitment: 0.25 | |
| codebook_loss_weight: 2.0 | |
| use_l2_normlize: True | |
| threshold_ema_dead_code: 0.2 | |
| speaker_encoder: | |
| input_dim: 128 | |
| out_dim: 1024 | |
| latent_dim: 128 | |
| token_num: 32 | |
| fsq_levels: [4, 4, 4, 4, 4, 4] | |
| fsq_num_quantizers: 1 | |
| prenet: | |
| input_channels: 1024 | |
| vocos_dim: 384 | |
| vocos_intermediate_dim: 2048 | |
| vocos_num_layers: 12 | |
| out_channels: 1024 | |
| condition_dim: 1024 | |
| sample_ratios: [1,1] | |
| use_tanh_at_final: False | |
| postnet: | |
| input_channels: 1024 | |
| vocos_dim: 384 | |
| vocos_intermediate_dim: 2048 | |
| vocos_num_layers: 6 | |
| out_channels: 1024 | |
| use_tanh_at_final: False | |
| highpass_cutoff_freq: 40 | |
| sample_rate: 16000 | |
| segment_duration: 2.4 # (s) | |
| max_val_duration: 12 # (s) | |
| latent_hop_length: 320 | |
| ref_segment_duration: 6 | |
| volume_normalize: true | |