Ukrainian TTS (text-to-speech) using Coqui TTS

Last update: Dec 26, 2022

Overview

title	emoji	colorFrom	colorTo	sdk	app_file	pinned
Ukrainian TTS	🐸	green	green	gradio	app.py	false

Ukrainian TTS 📢 🤖

Ukrainian TTS (text-to-speech) using Coqui TTS.

Trained on M-AILABS Ukrainian dataset using sumska voice.

Link to online demo -> https://huggingface.co/spaces/robinhad/ukrainian-tts

Support

If you like my work, please support -> SUPPORT LINK

Example

test.mp4

How to use :

pip install -r requirements.txt.
Download model from "Releases" tab.
Launch as one-time command:

tts --text "Text for TTS" \
    --model_path path/to/model.pth.tar \
    --config_path path/to/config.json \
    --out_path folder/to/save/output.wav

or alternatively launch web server using:

tts-server --model_path path/to/model.pth.tar \
    --config_path path/to/config.json

How to train:

Refer to "Nervous beginner guide" in Coqui TTS docs.
Instead of provided config.json use one from this repo.

Attribution

Code for app.py taken from https://huggingface.co/spaces/julien-c/coqui

Comments

Error with file: speakers.pth

FileNotFoundError: [Errno 2] No such file or directory: '/home/user/Soft/Python/mamba1/TTS/vits_mykyta_latest-September-12-2022_12+38AM-829e2c24/speakers.pth'

opened by akirsoft 4

doc: fix examples in README

Problem

The one-time snippet does not work as is and complains that the speaker is not defined

 > initialization of speaker-embedding layers.
 > Text: Перевірка мікрофона
 > Text splitted to sentences.
['Перевірка мікрофона']
Traceback (most recent call last):
  File "/home/serg/.local/bin/tts", line 8, in <module>
    sys.exit(main())
  File "/home/serg/.local/lib/python3.8/site-packages/TTS/bin/synthesize.py", line 350, in main
    wav = synthesizer.tts(
  File "/home/serg/.local/lib/python3.8/site-packages/TTS/utils/synthesizer.py", line 228, in tts
    raise ValueError(
ValueError:  [!] Look like you use a multi-speaker model. You need to define either a `speaker_name` or a `speaker_wav` to use a multi-speaker model.

Also it speakers.pth should be downloaded.

Fix

Just a few documentation changes:

make instructions on what to download from Releases more precise
add --speaker_id argument with one of the speakers

opened by seriar 2

One vowel words in the end of the sentence aren't stressed

Input:


Бобер на березі з бобренятами бублики пік.

Боронила борона по боронованому полю.

Ішов Прокіп, кипів окріп, прийшов Прокіп - кипить окріп, як при Прокопі, так і при Прокопі і при Прокопенятах.

Сидить Прокоп — кипить окроп, Пішов Прокоп — кипить окроп. Як при Прокопові кипів окроп, Так і без Прокопа кипить окроп.

Result:


Боб+ер н+а березі з бобрен+ятами б+ублики пік.

Борон+ила борон+а п+о борон+ованому п+олю.

Іш+ов Пр+окіп, кип+ів окр+іп, прийш+ов Пр+окіп - кип+ить окр+іп, +як пр+и Пр+окопі, т+ак +і пр+и Пр+окопі +і пр+и Прокопенятах.

Сид+ить Прок+оп — кип+ить окроп, Піш+ов Прок+оп — кип+ить окроп. +Як пр+и Пр+окопові кип+ів окроп, Т+ак +і б+ез Пр+окопа кип+ить окроп.```

opened by robinhad 0

Error import StressOption

Traceback (most recent call last): File "/home/user/Soft/Python/mamba1/test.py", line 1, in from ukrainian_tts.tts import TTS, Voices, StressOption ImportError: cannot import name 'StressOption' from 'ukrainian_tts.tts'

opened by akirsoft 0

Vits improvements

vitsArgs = VitsArgs(
    # hifi V3
    resblock_type_decoder = '2',
    upsample_rates_decoder = [8,8,4],
    upsample_kernel_sizes_decoder = [16,16,8],
    upsample_initial_channel_decoder = 256,
    resblock_kernel_sizes_decoder = [3,5,7],
    resblock_dilation_sizes_decoder = [[1,2], [2,6], [3,12]],
)

opened by robinhad 0

Model improvement checklist
[x] Add Ukrainian accentor - https://github.com/egorsmkv/ukrainian-accentor

[ ] Fine-tune from existing checkpoint (e.g. VITS Ljspeech)

[ ] Try to increase fft_size, hop_length to match sample_rate accordingly

[ ] Include more dataset samples into model
opened by robinhad 0

Releases(v4.0.0)

v4.0.0(Dec 10, 2022)

This is a release of Ukrainian TTS model and checkpoint. License for this model is GNU GPL v3 License. This release has a stress support using + sign before vowels. Model was trained for 160 000 steps by @robinhad .
Source code(tar.gz)
Source code(zip)
config.yaml(7.88 KB)
model.pth(368.17 MB)
v3.0.0(Sep 14, 2022)
This is a release of Ukrainian TTS model and checkpoint. License for this model is GNU GPL v3 License. This release has a stress support using + sign before vowels. Model was trained for 280 000 steps by @robinhad . Kudos to @egorsmkv for providing dataset for this model. Kudos to @proger for providing alignment scripts. Kudos to @dchaplinsky for Dmytro voice.

Example:

Test sentence:

К+ам'ян+ець-Под+ільський - м+істо в Хмельн+ицькій +області Укра+їни, ц+ентр Кам'ян+ець-Под+ільської міськ+ої об'+єднаної територі+альної гром+ади +і Кам'ян+ець-Под+ільського рай+ону.

Mykyta (male):

https://user-images.githubusercontent.com/5759207/190852232-34956a1d-77a9-42b9-b96d-39d0091e3e34.mp4

Olena (female):

https://user-images.githubusercontent.com/5759207/190852238-366782c1-9472-45fc-8fea-31346242f927.mp4

Dmytro (male):

https://user-images.githubusercontent.com/5759207/190852251-db105567-52ba-47b5-8ec6-5053c3baac8c.mp4

Olha (female):

https://user-images.githubusercontent.com/5759207/190852259-c6746172-05c4-4918-8286-a459c654eef1.mp4

Lada (female):

https://user-images.githubusercontent.com/5759207/190852270-7aed2db9-dc08-4a9f-8775-07b745657ca1.mp4
Source code(tar.gz)
Source code(zip)
config.json(12.07 KB)
model-inference.pth(329.95 MB)
model.pth(989.97 MB)
speakers.pth(495 bytes)
v3.0.0-alpha(Sep 8, 2022)

Mykyta, Lada and Olena License: GPL v3 licence.
Source code(tar.gz)
Source code(zip)
config.json(9.94 KB)
model-inference.pth(329.95 MB)
model.pth(989.96 MB)
speakers.pth(431 bytes)
v2.0.0(Jul 10, 2022)
This is a release of Ukrainian TTS model and checkpoint using voice (7 hours) from Mykyta dataset. License for this model is GNU GPL v3 License. This release has a stress support using + sign before vowels. Model was trained for 140 000 steps by @robinhad . Kudos to @egorsmkv for providing Mykyta and Olena dataset.

Example:

Test sentence:

К+ам'ян+ець-Под+ільський - м+істо в Хмельн+ицькій +області Укра+їни, ц+ентр Кам'ян+ець-Под+ільської міськ+ої об'+єднаної територі+альної гром+ади +і Кам'ян+ець-Под+ільського рай+ону.

Mykyta (male):

https://user-images.githubusercontent.com/5759207/178158485-29a5d496-7eeb-4938-8ea7-c345bc9fed57.mp4

Olena (female):

https://user-images.githubusercontent.com/5759207/178158492-8504080e-2f13-43f1-83f0-489b1f9cd66b.mp4
Source code(tar.gz)
Source code(zip)
config.json(9.97 KB)
model-inference.pth(329.95 MB)
model.pth(989.72 MB)
optimized.pth(329.95 MB)
speakers.pth(431 bytes)
v2.0.0-beta(May 8, 2022)

This is a beta release of Ukrainian TTS model and checkpoint using voice (7 hours) from Mykyta dataset. License for this model is GNU GPL v3 License. This release has a stress support using + sign before vowels. Model was trained for 150 000 steps by @robinhad . Kudos to @egorsmkv for providing Mykyta dataset.

Example:

https://user-images.githubusercontent.com/5759207/167305810-2b023da7-0657-44ac-961f-5abf1aa6ea7d.mp4

:
Source code(tar.gz)
Source code(zip)
config.json(8.85 KB)
LICENSE(34.32 KB)
model-inference.pth(317.15 MB)
model.pth(951.32 MB)
tts_output.wav(1.11 MB)
v1.0.0(Jan 14, 2022)

This is release of Ukrainian TTS model and checkpoint using voice (10 hours) from M-AILABS Ukrainian dataset. Model was trained using 200 000 steps. Example:

https://user-images.githubusercontent.com/5759207/149566245-40656002-3999-48a8-b671-e0f74c3d6e2f.mp4
Source code(tar.gz)
Source code(zip)
config.json(6.44 KB)
example.mp4(32.07 KB)
v0.0.1(Oct 14, 2021)

This is release of Ukrainian TTS model and checkpoint using voice (10 hours) from M-AILABS Ukrainian dataset. Model was trained using 14145 steps. Example:

https://user-images.githubusercontent.com/5759207/140622395-9e734c95-159c-4d72-9f56-e8d1f1ac66c2.mp4
Source code(tar.gz)
Source code(zip)
config.json(4.96 KB)
test.mp4(32.73 KB)
vocoder_config.json(4.39 KB)

Owner

Yurii Paniv

AI and stuff

GitHub Repository https://huggingface.co/spaces/robinhad/ukrainian-tts

Baseline code for Korean open domain question answering(ODQA)

Open-Domain Question Answering(ODQA)는 다양한 주제에 대한 문서 집합으로부터 자연어 질의에 대한 답변을 찾아오는 task입니다. 이때 사용자 질의에 답변하기 위해 주어지는 지문이 따로 존재하지 않습니다. 따라서 사전에 구축되어있는 Knowl

69 Nov 04, 2022

Utilize Korean BERT model in sentence-transformers library

ko-sentence-transformers 이 프로젝트는 KoBERT 모델을 sentence-transformers 에서 보다 쉽게 사용하기 위해 만들어졌습니다. Ko-Sentence-BERT-SKTBERT 프로젝트에서는 KoBERT 모델을 sentence-trans

40 Dec 20, 2022

BPEmb is a collection of pre-trained subword embeddings in 275 languages, based on Byte-Pair Encoding (BPE) and trained on Wikipedia.

BPEmb is a collection of pre-trained subword embeddings in 275 languages, based on Byte-Pair Encoding (BPE) and trained on Wikipedia. Its intended use is as input for neural models in natural languag

1.1k Jan 03, 2023

Ukrainian TTS (text-to-speech) using Coqui TTS

Related tags

Overview

Ukrainian TTS 📢 🤖

Support

Example

How to use :

How to train:

Attribution

Comments

Error with file: speakers.pth

doc: fix examples in README

Problem

Fix

One vowel words in the end of the sentence aren't stressed

Error import StressOption

Vits improvements

Model improvement checklist

Releases(v4.0.0)

v4.0.0(Dec 10, 2022)

v3.0.0(Sep 14, 2022)

Example:

v3.0.0-alpha(Sep 8, 2022)

v2.0.0(Jul 10, 2022)

Example:

v2.0.0-beta(May 8, 2022)

Example:

v1.0.0(Jan 14, 2022)

v0.0.1(Oct 14, 2021)

Owner

Yurii Paniv

Baseline code for Korean open domain question answering(ODQA)

Utilize Korean BERT model in sentence-transformers library

BPEmb is a collection of pre-trained subword embeddings in 275 languages, based on Byte-Pair Encoding (BPE) and trained on Wikipedia.

Reproduction process of BERT on SST2 dataset

A look-ahead multi-entity Transformer for modeling coordinated agents.

Pervasive Attention: 2D Convolutional Networks for Sequence-to-Sequence Prediction

Trains an OpenNMT PyTorch model and SentencePiece tokenizer.

BERTAC (BERT-style transformer-based language model with Adversarially pretrained Convolutional neural network)

A Python 3.6+ package to run .many files, where many programs written in many languages may exist in one file.

Continuously update some NLP practice based on different tasks.

This repository contains helper functions which can help you generate additional data points depending on your NLP task.

Paddlespeech Streaming ASR GUI

This is a general repo that helps you develop fast/effective NLP classifiers using Huggingface

Parrot is a paraphrase based utterance augmentation framework purpose built to accelerate training NLU models

A paper list for aspect based sentiment analysis.

Easy Language Model Pretraining leveraging Huggingface's Transformers and Datasets

Code for text augmentation method leveraging large-scale language models

Stand-alone language identification system

A programming language with logic of Python, and syntax of all languages.

Unlimited Call - Text Bombing Tool