TheShadow29 / Vc With Gan
Voice Conversion with GANs
Stars: ✭ 13
Programming Languages
python
139335 projects - #7 most used programming language
Labels
Projects that are alternatives of or similar to Vc With Gan
Segan
A PyTorch implementation of SEGAN based on INTERSPEECH 2017 paper "SEGAN: Speech Enhancement Generative Adversarial Network"
Stars: ✭ 82 (+530.77%)
Mutual labels: gan, voice
Inverse Style Gan
Looking up a generative latent vectors from (face) reference images.
Stars: ✭ 26 (+100%)
Mutual labels: gan
Csmri Refinement
Code for "Adversarial and Perceptual Refinement Compressed Sensing MRI Reconstruction"
Stars: ✭ 21 (+61.54%)
Mutual labels: gan
Instagan
InstaGAN: Instance-aware Image Translation (ICLR 2019)
Stars: ✭ 761 (+5753.85%)
Mutual labels: gan
Transgan
[Preprint] "TransGAN: Two Transformers Can Make One Strong GAN", Yifan Jiang, Shiyu Chang, Zhangyang Wang
Stars: ✭ 864 (+6546.15%)
Mutual labels: gan
Say
Convert text to audiable speech. Play it or save it to audio file.
Stars: ✭ 24 (+84.62%)
Mutual labels: voice
Began Tensorflow
Tensorflow implementation of "BEGAN: Boundary Equilibrium Generative Adversarial Networks"
Stars: ✭ 904 (+6853.85%)
Mutual labels: gan
Xdf Gan
A GAN for the generation of mock astronomical surveys
Stars: ✭ 17 (+30.77%)
Mutual labels: gan
Tensorlayer
Deep Learning and Reinforcement Learning Library for Scientists and Engineers 🔥
Stars: ✭ 6,796 (+52176.92%)
Mutual labels: gan
Bernard
Bernard is a voice assistant developed with gTTS. It can fulfill basic and simple tasks you give.
Stars: ✭ 24 (+84.62%)
Mutual labels: voice
Generative Models
Collection of generative models, e.g. GAN, VAE in Pytorch and Tensorflow.
Stars: ✭ 6,701 (+51446.15%)
Mutual labels: gan
Xunfei Clj
Clojure封装讯飞语音SDK, 可提供给Emacs/Vim编辑器使用,或者命令行, 实现语音提醒/语音识别/语音转为命令等
Stars: ✭ 26 (+100%)
Mutual labels: voice
Pytorch Pretrained Biggan
🦋A PyTorch implementation of BigGAN with pretrained weights and conversion scripts.
Stars: ✭ 779 (+5892.31%)
Mutual labels: gan
Lightning Bolts
Toolbox of models, callbacks, and datasets for AI/ML researchers.
Stars: ✭ 829 (+6276.92%)
Mutual labels: gan
Advanced Deep Learning With Keras
Advanced Deep Learning with Keras, published by Packt
Stars: ✭ 917 (+6953.85%)
Mutual labels: gan
St Cgan
Dataset and Code for our CVPR'18 paper ST-CGAN: "Stacked Conditional Generative Adversarial Networks for Jointly Learning Shadow Detection and Shadow Removal"
Stars: ✭ 13 (+0%)
Mutual labels: gan
Vonage Php Sdk Core
Vonage REST API client for PHP. API support for SMS, Voice, Text-to-Speech, Numbers, Verify (2FA) and more.
Stars: ✭ 849 (+6430.77%)
Mutual labels: voice
VC-with-GAN
CS 753 ASR project
Usage Steps:
- Run
bash download.sh
to prepare the VCC2018 dataset. - Run
analyzer.py
to extract features and write features into binary files. (This takes a few minutes.) - Run
build.py
to record some stats, such as spectral extrema and pitch. - To train a VAWGAN, for example, run
python main.py \
--model VAWGAN \
--trainer VAWGANTrainer \
--architecture architecture-vawgan-vcc2016.json
- You can find your models in
./logdir/train/[timestamp]
- To convert the voice, run
python convert.py \
--src VCC2SF1 \
--trg VCC2TM1 \
--model VAWGAN \
--checkpoint logdir/train/[timestamp]/[model.ckpt-[id]] \
--file_pattern "./dataset/vcc2018/bin/Training Set/{}/[0-9]*.bin"
*Please fill in timestamp
and model id
.
7. You can find the converted wav files in ./logdir/output/[timestamp]
8. If you want to convert all the voices, run
./convert_all.sh \
--model VAWGAN \
--checkpoint logdir/train/[timestamp]/[model.ckpt-[id]] \
--output_dir [directory to store converted audio]
Usage for Sentence Embeddings:
-
Ensure you have
w_prob_dict.pkl
andw_vec_dict.pkl
indata
directory.- For
w_prob_dict.pkl
you have two options. Either useget_word_prob_from_corpus
this demands a corpus as an input. We used WikiText. Or you can get a csv file with unigram probabilities (we mentioned the source in the report http://norvig.com/ngrams/) and use the function get_w_prob_from_csv. - For
w_vec_dict.pkl initialize
a Sentence_Embedding object and then call the functionprune_word_vec
. This essentially keeps only those embeddings which are present in the transcriptions since it takes a lot more time (and ram) to get the parse the whole fasttext data. - All pickle files are shared here https://drive.google.com/drive/folders/1FWGGEQ9wTUewBDFq5ssT4BP4cMyt8lh1
- For
-
Download the dataset using
bash download.sh
-
Run
python sentence_embedding.py
. This should createsent_emb.pkl
insidedata
directory. -
Run
analyzer.py
to extract features, store them along with sentence embeddings. -
Run
build.py
to find statistics about features. -
To train with sentence embedding, run
python main.py \
--model VAWGAN_S \
--trainer VAWGAN_S \
--architecture architecture-vawgan-sent.json
- For conversion, run
python convert.py \
--src VCC2SF1 \
--trg VCC2TM1 \
--model VAWGAN_S \
--checkpoint logdir/train/[timestamp]/[model.ckpt-[id]] \
--file_pattern "./dataset/vcc2018/bin/Training Set/{}/[0-9]*.bin"
or
./convert_all.sh \
--model VAWGAN_S \
--checkpoint logdir/train/[timestamp]/[model.ckpt-[id]] \
--output_dir [directory to store converted audio]
Note that the project description data, including the texts, logos, images, and/or trademarks,
for each open source project belongs to its rightful owner.
If you wish to add or remove any projects, please contact us at [email protected].