%20--%3e%3cdefs%3e%3cstyle%3e%20.st0%20{%20fill:%20%23061b40;%20}%20.st1%20{%20fill:%20%23306af1;%20}%20.st2%20{%20fill:%20%235ce5cf;%20}%20%3c/style%3e%3c/defs%3e%3cg%3e%3cpath%20class='st0'%20d='M55,10.5h9v3h-9v9h12V7.5h-12v3ZM64,19.5h-6v-3h6v3Z'/%3e%3cpolygon%20class='st0'%20points='69%2016.5%2078%2016.5%2078%2019.5%2069%2019.5%2069%2022.5%2081%2022.5%2081%2013.5%2072%2013.5%2072%2010.5%2081%2010.5%2081%207.5%2069%207.5%2069%2016.5'/%3e%3cpolygon%20class='st0'%20points='95%2010.5%2095%207.5%2083%207.5%2083%2022.5%2095%2022.5%2095%2019.5%2086%2019.5%2086%2016.5%2095%2016.5%2095%2013.5%2086%2013.5%2086%2010.5%2095%2010.5'/%3e%3cpath%20class='st0'%20d='M40,1.5v21h11.6l1.4-1.4v-7.6h0c0,0-1.4-1.5-1.4-1.5l1.4-1.4V2.9l-1.4-1.4h-11.6ZM50,19.5h-7v-6h7v6ZM50,10.5h-7v-6h7v6Z'/%3e%3c/g%3e%3cpath%20class='st1'%20d='M23.1,24L14.7,4.8l-4.9,11.2h3.8l-1.8,4H3.3L12.1,0H2C.9,0,0,.9,0,2v20c0,1.1.9,2,2,2h21.1Z'/%3e%3cpath%20class='st2'%20d='M34,0h-16.8l10.6,24h6.2c1.1,0,2-.9,2-2V2C36,.9,35.1,0,34,0ZM32.5,20h-4V4h4v16Z'/%3e%3c/svg%3e)
EN
Gunnarthor Talromur B Fastspeech2
Icelandic text-to-speech model based on FastSpeech2 architecture, trained using the talromur dataset
Downloads 53
Release Time : 4/19/2022
Model Overview
This is a FastSpeech2 text-to-speech model based on the ESPnet framework, specifically optimized for Icelandic, capable of converting text into natural speech output.
Model Features
FastSpeech2 Architecture
Utilizes the FastSpeech2 model architecture, providing efficient speech synthesis capabilities
Icelandic Optimization
Specifically trained and optimized for the Icelandic language
Pitch and Energy Prediction
Includes independent pitch and energy prediction modules to enhance speech naturalness
Model Capabilities
Text-to-Speech
Icelandic Speech Synthesis
Pitch Control
Energy Control
Use Cases
Speech Synthesis
Audiobook Generation
Convert Icelandic text into natural speech for audiobooks
Voice Assistants
Provide speech synthesis capabilities for Icelandic voice assistants
🚀 ESPnet2 TTS model
This is an ESPnet2-based TTS model, trained by Gunnar Thor using the talromur recipe, providing high - quality text - to - speech conversion.
🚀 Quick Start
📦 Installation
To use this model in ESPnet2, follow these steps:
cd espnet
git checkout 49a284e69308d81c142b89795de255b4ce290c54
pip install -e .
cd egs2/talromur/tts1
./run.sh --skip_data_prep false --skip_train true --download_model espnet/GunnarThor_talromur_b_fastspeech2
✨ Features
- Trained by Expert: This model was trained by Gunnar Thor using the talromur recipe in espnet.
- Easy to Use: With clear installation steps, it can be quickly integrated into ESPnet2 projects.
📚 Documentation
TTS config
expand
config: conf/tuning/train_fastspeech2.yaml
print_config: false
log_level: INFO
dry_run: false
iterator_type: sequence
output_dir: exp/b/tts_train_fastspeech2_raw_phn_none
ngpu: 1
seed: 0
num_workers: 1
num_att_plot: 3
dist_backend: nccl
dist_init_method: env://
dist_world_size: null
dist_rank: null
local_rank: 0
dist_master_addr: null
dist_master_port: null
dist_launcher: null
multiprocessing_distributed: false
unused_parameters: false
sharded_ddp: false
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: true
collect_stats: false
write_collected_feats: false
max_epoch: 100
patience: null
val_scheduler_criterion:
- valid
- loss
early_stopping_criterion:
- valid
- loss
- min
best_model_criterion:
- - valid
- loss
- min
- - train
- loss
- min
keep_nbest_models: 5
nbest_averaging_interval: 0
grad_clip: 1.0
grad_clip_type: 2.0
grad_noise: false
accum_grad: 8
no_forward_run: false
resume: true
train_dtype: float32
use_amp: false
log_interval: null
use_matplotlib: true
use_tensorboard: true
use_wandb: false
wandb_project: null
wandb_id: null
wandb_entity: null
wandb_name: null
wandb_model_log_interval: -1
detect_anomaly: false
pretrain_path: null
init_param: []
ignore_init_mismatch: false
freeze_param: []
num_iters_per_epoch: 800
batch_size: 20
valid_batch_size: null
batch_bins: 2500000
valid_batch_bins: null
train_shape_file:
- exp/b/tts_train_tacotron2_raw_phn_none/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/text_shape.phn
- exp/b/tts_train_tacotron2_raw_phn_none/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/speech_shape
valid_shape_file:
- exp/b/tts_train_tacotron2_raw_phn_none/decode_use_teacher_forcingtrue_train.loss.ave/stats/valid/text_shape.phn
- exp/b/tts_train_tacotron2_raw_phn_none/decode_use_teacher_forcingtrue_train.loss.ave/stats/valid/speech_shape
batch_type: numel
valid_batch_type: null
fold_length:
- 150
- 204800
sort_in_batch: descending
sort_batch: descending
multiple_iterator: false
chunk_length: 500
chunk_shift_ratio: 0.5
num_cache_chunks: 1024
train_data_path_and_name_and_type:
- - dump/raw/train_b_phn/text
- text
- text
- - exp/b/tts_train_tacotron2_raw_phn_none/decode_use_teacher_forcingtrue_train.loss.ave/train_b_phn/durations
- durations
- text_int
- - dump/raw/train_b_phn/wav.scp
- speech
- sound
valid_data_path_and_name_and_type:
- - dump/raw/dev_b_phn/text
- text
- text
- - exp/b/tts_train_tacotron2_raw_phn_none/decode_use_teacher_forcingtrue_train.loss.ave/dev_b_phn/durations
- durations
- text_int
- - dump/raw/dev_b_phn/wav.scp
- speech
- sound
allow_variable_data_keys: false
max_cache_size: 0.0
max_cache_fd: 32
valid_max_cache_size: null
optim: adam
optim_conf:
lr: 1.0
scheduler: noamlr
scheduler_conf:
model_size: 384
warmup_steps: 4000
token_list:
- <blank>
- <unk>
- ','
- .
- r
- t
- n
- a0
- s
- I0
- D
- l
- Y0
- m
- v
- h
- E1
- k
- a:1
- E:1
- j
- f
- T
- G
- a1
- p
- c
- au:1
- i:1
- O:1
- E0
- I:1
- r_0
- I1
- t_h
- k_h
- Y1
- i0
- ei1
- u:1
- ou:1
- ei:1
- O1
- N
- l_0
- '91'
- ou0
- ai0
- n_0
- au1
- O0
- ou1
- ai:1
- ei0
- '9:1'
- ai1
- i1
- c_h
- '90'
- au0
- x
- C
- p_h
- u0
- 9i:1
- Y:1
- 9i1
- J
- u1
- 9i0
- N_0
- m_0
- J_0
- Oi1
- Yi0
- Yi1
- Oi0
- '9:0'
- au:0
- E:0
- <sos/eos>
odim: null
model_conf: {}
use_preprocessor: true
token_type: phn
bpemodel: null
non_linguistic_symbols: null
cleaner: null
g2p: null
feats_extract: fbank
feats_extract_conf:
n_fft: 1024
hop_length: 256
win_length: null
fs: 22050
fmin: 80
fmax: 7600
n_mels: 80
normalize: global_mvn
normalize_conf:
stats_file: exp/b/tts_train_tacotron2_raw_phn_none/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/feats_stats.npz
tts: fastspeech2
tts_conf:
adim: 384
aheads: 2
elayers: 4
eunits: 1536
dlayers: 4
dunits: 1536
positionwise_layer_type: conv1d
positionwise_conv_kernel_size: 3
duration_predictor_layers: 2
duration_predictor_chans: 256
duration_predictor_kernel_size: 3
postnet_layers: 5
postnet_filts: 5
postnet_chans: 256
use_masking: true
use_scaled_pos_enc: true
encoder_normalize_before: true
decoder_normalize_before: true
reduction_factor: 1
init_type: xavier_uniform
init_enc_alpha: 1.0
init_dec_alpha: 1.0
transformer_enc_dropout_rate: 0.2
transformer_enc_positional_dropout_rate: 0.2
transformer_enc_attn_dropout_rate: 0.2
transformer_dec_dropout_rate: 0.2
transformer_dec_positional_dropout_rate: 0.2
transformer_dec_attn_dropout_rate: 0.2
pitch_predictor_layers: 5
pitch_predictor_chans: 256
pitch_predictor_kernel_size: 5
pitch_predictor_dropout: 0.5
pitch_embed_kernel_size: 1
pitch_embed_dropout: 0.0
stop_gradient_from_pitch_predictor: true
energy_predictor_layers: 2
energy_predictor_chans: 256
energy_predictor_kernel_size: 3
energy_predictor_dropout: 0.5
energy_embed_kernel_size: 1
energy_embed_dropout: 0.0
stop_gradient_from_energy_predictor: false
pitch_extract: dio
pitch_extract_conf:
fs: 22050
n_fft: 1024
hop_length: 256
f0max: 400
f0min: 80
reduction_factor: 1
pitch_normalize: global_mvn
pitch_normalize_conf:
stats_file: exp/b/tts_train_tacotron2_raw_phn_none/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/pitch_stats.npz
energy_extract: energy
energy_extract_conf:
fs: 22050
n_fft: 1024
hop_length: 256
win_length: null
reduction_factor: 1
energy_normalize: global_mvn
energy_normalize_conf:
stats_file: exp/b/tts_train_tacotron2_raw_phn_none/decode_use_teacher_forcingtrue_train.loss.ave/stats/train/energy_stats.npz
required:
- output_dir
- token_list
version: 0.10.7a1
distributed: false
Citing ESPnet
@inproceedings{watanabe2018espnet,
author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
title={{ESPnet}: End-to-End Speech Processing Toolkit},
year={2018},
booktitle={Proceedings of Interspeech},
pages={2207--2211},
doi={10.21437/Interspeech.2018-1456},
url={http://dx.doi.org/10.21437/Interspeech.2018-1456}
}
@inproceedings{hayashi2020espnet,
title={{Espnet-TTS}: Unified, reproducible, and integratable open source end-to-end text-to-speech toolkit},
author={Hayashi, Tomoki and Yamamoto, Ryuichi and Inoue, Katsuki and Yoshimura, Takenori and Watanabe, Shinji and Toda, Tomoki and Takeda, Kazuya and Zhang, Yu and Tan, Xu},
booktitle={Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
pages={7654--7658},
year={2020},
organization={IEEE}
}
or arXiv:
@misc{watanabe2018espnet,
title={ESPnet: End-to-End Speech Processing Toolkit},
author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
year={2018},
eprint={1804.00015},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
📄 License
This project is licensed under the CC - BY - 4.0 license.
Kokoro 82M
Apache-2.0
Kokoro is an open-source text-to-speech (TTS) model with 82 million parameters, renowned for its lightweight architecture and high audio quality, while also being fast and cost-effective.
Speech Synthesis English
K
hexgrad
2.0M
4,155
XTTS V2
Other
ⓍTTS is a revolutionary voice generation model that achieves cross-lingual voice cloning with just a 6-second audio clip, supporting 17 languages.
Speech Synthesis
X
coqui
1.7M
2,630
F5 TTS
F5-TTS is a flow matching-based voice synthesis model, focusing on fluent and faithful voice synthesis, especially suitable for scenarios like fairy tale narration.
Speech Synthesis
F
SWivid
851.49k
1,000
Bigvgan V2 22khz 80band 256x
MIT
BigVGAN is a general-purpose neural vocoder trained at scale, capable of generating high-quality audio waveforms from mel spectrograms.
Speech Synthesis
B
nvidia
503.23k
16
Speecht5 Tts
MIT
A SpeechT5 speech synthesis (text-to-speech) model fine-tuned on the LibriTTS dataset, supporting high-quality text-to-speech conversion.
Speech Synthesis
Transformers
TransformersS
microsoft
113.83k
760
Dia 1.6B
Apache-2.0
Dia is a 1.6 billion parameter text-to-speech model developed by Nari Labs, capable of generating highly realistic conversations directly from text, supporting emotional and tonal control, and producing non-verbal communication content.
Speech Synthesis
Safetensors English
Safetensors EnglishD
nari-labs
80.28k
1,380
Csm 1b
Apache-2.0
CSM is a 1-billion-parameter voice generation model developed by Sesame, capable of generating RVQ audio encoding from text and audio inputs
Speech Synthesis Safetensors English
Safetensors EnglishC
sesame
65.03k
1,950
Kokoro 82M V1.1 Zh
Apache-2.0
Kokoro is an open-weight series of small yet powerful text-to-speech (TTS) models, now featuring data from 100 Chinese speakers sourced from professional datasets.
Speech Synthesis
K
hexgrad
51.56k
112
Indic Parler Tts
Apache-2.0
Indic Parler-TTS is a multilingual extension of Parler-TTS Mini, supporting 21 languages including various Indian languages and English.
Speech Synthesis Transformers Supports Multiple Languages
Transformers Supports Multiple LanguagesI
ai4bharat
43.59k
124
Bark
MIT
Bark is a Transformer-based text-to-audio model created by Suno, capable of generating highly realistic multilingual speech, music, background noise, and simple sound effects.
Speech Synthesis Transformers Supports Multiple Languages
Transformers Supports Multiple LanguagesB
suno
35.72k
1,326
Featured Recommended AI Models