Code

🤖 AUTO TTS #696

Code

Closed

loganhart02 wants to merge 132 commits intocoqui-ai:devcoqui-ai/TTS:devfrom loganhart02:recipe_apiCopy head branch name to clipboard

+1,353-1

Commits

Commits on Jul 27, 2021

• Skip phoneme cache pre-compute if the path exists

  erogol

  committed
  88d782e

  Copy full SHA for 88d782e

Commits on Aug 2, 2021

• Fix json formatting error

  Moldoteck

  authored
  66bcb8f

  Copy full SHA for 66bcb8f

• Update distribute.py (#685)

  Show description for cf55494
  astricks

  authored
  cf55494

  Copy full SHA for cf55494

• Merge pull request #694 from Moldoteck/patch-1

  Show description for 168a9a6
  erogol

  authored
  168a9a6

  Copy full SHA for 168a9a6

Commits on Aug 3, 2021

• nice way to train complete recipes with three lines of code. currently only made ljspeech recipes

  loganhart02

  committed
  c6e5121

  Copy full SHA for c6e5121

Commits on Aug 6, 2021

• added command line training file for ljspeech tacotron2 models.

  loganhart02

  committed
  8c2a2c5

  Copy full SHA for 8c2a2c5

• added SCGlowTts to model hub

  loganhart02

  committed
  1de4995

  Copy full SHA for 1de4995

Commits on Aug 8, 2021

• added recipe for sc glow tts training on vctk and tacotron2 models for sam dataset, also added options so users can turn on forward and loction attention when they want to.

  loganhart02

  committed
  55fbcdb

  Copy full SHA for 55fbcdb

Commits on Aug 12, 2021

• hifi gan and wavegrad configs for and ljspeech vocoder recipe

  loganhart02

  committed
  42bd0dc

  Copy full SHA for 42bd0dc

Commits on Aug 14, 2021

• tacotron2 config

  loganhart02

  committed
  b84301d

  Copy full SHA for b84301d

Commits on Aug 15, 2021

• Merge branch 'main' of https://github.com/coqui-ai/TTS into recipe_api

  loganhart02

  committed
  ae22b23

  Copy full SHA for ae22b23

• Merge branch 'coqui-ai:main' into recipe_api

  loganhart02

  authored
  7d9029d

  Copy full SHA for 7d9029d

• Merge remote-tracking branch 'origin/recipe_api' into recipe_api

  loganhart02

  committed
  807d432

  Copy full SHA for 807d432

• made ljspeech tts trainer for all models( define models by string instead of calling a function) added more ljspeech vocoder models(these are just baselines though, once I have a free gpu im going …

  Show description for 42fb3fd
  loganhart02

  committed
  42fb3fd

  Copy full SHA for 42fb3fd

Commits on Aug 16, 2021

• made ljspeech tts and vocoder trainer.

  loganhart02

  committed
  55d9fc3

  Copy full SHA for 55d9fc3

Commits on Aug 19, 2021

• vocoder command line file

  loganhart02

  committed
  68975bd

  Copy full SHA for 68975bd

Commits on Aug 23, 2021

• added vits(forgot to push this in my last commit)

  loganhart02

  committed
  dd80696

  Copy full SHA for dd80696

Commits on Sep 1, 2021

• added vits default model

  loganhart02

  committed
  88e2e39

  Copy full SHA for 88e2e39

• Merge branch 'coqui-ai:main' into recipe_api

  loganhart02

  authored
  e5c3c27

  Copy full SHA for e5c3c27

Commits on Sep 2, 2021

• nice way to train complete recipes with three lines of code. currently only made ljspeech recipes

  loganhart02

  committed
  e4b08ce

  Copy full SHA for e4b08ce

• added command line training file for ljspeech tacotron2 models.

  loganhart02

  committed
  391348d

  Copy full SHA for 391348d

• added SCGlowTts to model hub

  loganhart02

  committed
  d4f1551

  Copy full SHA for d4f1551

• added recipe for sc glow tts training on vctk and tacotron2 models for sam dataset, also added options so users can turn on forward and loction attention when they want to.

  loganhart02

  committed
  71e9ea3

  Copy full SHA for 71e9ea3

• hifi gan and wavegrad configs for and ljspeech vocoder recipe

  loganhart02

  committed
  17a8dd6

  Copy full SHA for 17a8dd6

• tacotron2 config

  loganhart02

  committed
  6f0ab09

  Copy full SHA for 6f0ab09

• Update default ja vocoder

  

  erogol

  authored and

  loganhart02

  committed
  3a16744

  Copy full SHA for 3a16744

• made ljspeech tts trainer for all models( define models by string instead of calling a function) added more ljspeech vocoder models(these are just baselines though, once I have a free gpu im going …

  Show description for bc967eb
  loganhart02

  committed
  bc967eb

  Copy full SHA for bc967eb

• made ljspeech tts and vocoder trainer.

  loganhart02

  committed
  3af309e

  Copy full SHA for 3af309e

• vocoder command line file

  loganhart02

  committed
  fb4cf15

  Copy full SHA for fb4cf15

• added vits(forgot to push this in my last commit)

  loganhart02

  committed
  fe301dc

  Copy full SHA for fe301dc

• Merge branch 'recipe_api' of https://github.com/loganhart420/TTS into recipe_api

  loganhart02

  committed
  3e862d8

  Copy full SHA for 3e862d8

Commits on Sep 6, 2021

• Merge branch 'coqui-ai:main' into recipe_api

  loganhart02

  authored
  6bbd811

  Copy full SHA for 6bbd811

• nice way to train complete recipes with three lines of code. currently only made ljspeech recipes

  loganhart02

  committed
  c5d2334

  Copy full SHA for c5d2334

• added command line training file for ljspeech tacotron2 models.

  loganhart02

  committed
  51380ec

  Copy full SHA for 51380ec

• added recipe for sc glow tts training on vctk and tacotron2 models for sam dataset, also added options so users can turn on forward and loction attention when they want to.

  loganhart02

  committed
  4275b96

  Copy full SHA for 4275b96

• hifi gan and wavegrad configs for and ljspeech vocoder recipe

  loganhart02

  committed
  e47f763

  Copy full SHA for e47f763

• made ljspeech tts trainer for all models( define models by string instead of calling a function) added more ljspeech vocoder models(these are just baselines though, once I have a free gpu im going …

  Show description for a7f1ac0
  loganhart02

  committed
  a7f1ac0

  Copy full SHA for a7f1ac0

• made ljspeech tts and vocoder trainer.

  loganhart02

  committed
  2cdaf08

  Copy full SHA for 2cdaf08

• vocoder command line file

  loganhart02

  committed
  0b5f397

  Copy full SHA for 0b5f397

• changed camel case to _. created single speaker and multispeaker auto tts functions

  loganhart02

  committed
  40b9eaa

  Copy full SHA for 40b9eaa

Commits on Sep 7, 2021

• made single speaker vocoder trainer function

  loganhart02

  committed
  9155412

  Copy full SHA for 9155412

• nice way to train complete recipes with three lines of code. currently only made ljspeech recipes

  loganhart02

  committed
  95e978d

  Copy full SHA for 95e978d

• added command line training file for ljspeech tacotron2 models.

  loganhart02

  committed
  190b308

  Copy full SHA for 190b308

• added SCGlowTts to model hub

  loganhart02

  committed
  5aa9002

  Copy full SHA for 5aa9002

• added recipe for sc glow tts training on vctk and tacotron2 models for sam dataset, also added options so users can turn on forward and loction attention when they want to.

  loganhart02

  committed
  fac98ca

  Copy full SHA for fac98ca

• hifi gan and wavegrad configs for and ljspeech vocoder recipe

  loganhart02

  committed
  ff2f974

  Copy full SHA for ff2f974

• tacotron2 config

  loganhart02

  committed
  ed38be6

  Copy full SHA for ed38be6

• Update default ja vocoder

  

  erogol

  authored and

  loganhart02

  committed
  68f7357

  Copy full SHA for 68f7357

• made ljspeech tts trainer for all models( define models by string instead of calling a function) added more ljspeech vocoder models(these are just baselines though, once I have a free gpu im going …

  Show description for 8ab2967
  loganhart02

  committed
  8ab2967

  Copy full SHA for 8ab2967

• made ljspeech tts and vocoder trainer.

  loganhart02

  committed
  d6f5da8

  Copy full SHA for d6f5da8

• vocoder command line file

  loganhart02

  committed
  485daec

  Copy full SHA for 485daec

• added vits(forgot to push this in my last commit)

  loganhart02

  committed
  e82dc31

  Copy full SHA for e82dc31

• nice way to train complete recipes with three lines of code. currently only made ljspeech recipes

  loganhart02

  committed
  7b67e79

  Copy full SHA for 7b67e79

• Update Japanese phonemizer (#758)

  Show description for d452288
  

  2 peopleauthored and

  loganhart02

  committed
  d452288

  Copy full SHA for d452288

• Compute F0 using librosa

  

  erogol

  authored and

  loganhart02

  committed
  85cfbe6

  Copy full SHA for 85cfbe6

• Add FastPitchLoss

  

  erogol

  authored and

  loganhart02

  committed
  cfcba2a

  Copy full SHA for cfcba2a

• Add comput_f0 field

  

  erogol

  authored and

  loganhart02

  committed
  1b67fe2

  Copy full SHA for 1b67fe2

• Fix compute_attention_masks.py

  

  erogol

  authored and

  loganhart02

  committed
  f3a3893

  Copy full SHA for f3a3893

• Fix configs

  

  erogol

  authored and

  loganhart02

  committed
  5f7f383

  Copy full SHA for 5f7f383

• Fix FastPitchLoss

  

  erogol

  authored and

  loganhart02

  committed
  ac9ac9b

  Copy full SHA for ac9ac9b

• Fix base_tacotron aux_input handling

  

  erogol

  authored and

  loganhart02

  committed
  bda1409

  Copy full SHA for bda1409

• Cache pitch features

  Show description for 3af4eda
  

  erogol

  authored and

  loganhart02

  committed
  3af4eda

  Copy full SHA for 3af4eda

• Set BaseDatasetConfig for tests

  

  erogol

  authored and

  loganhart02

  committed
  0da267b

  Copy full SHA for 0da267b

• Don't print computed phonemes

  Show description for 43b3ff7
  

  erogol

  authored and

  loganhart02

  committed
  43b3ff7

  Copy full SHA for 43b3ff7

• Compute mean and std pitch

  

  erogol

  authored and

  loganhart02

  committed
  306de95

  Copy full SHA for 306de95

• Add FastPitch LJSpeech recipe

  

  erogol

  authored and

  loganhart02

  committed
  a478f68

  Copy full SHA for a478f68

• Add yin based pitch computation

  

  erogol

  authored and

  loganhart02

  committed
  8978751

  Copy full SHA for 8978751

• Add FastPitch model and FastPitchconfig

  

  erogol

  authored and

  loganhart02

  committed
  5a4e98a

  Copy full SHA for 5a4e98a

• Use absolute paths of the attention masks

  

  erogol

  authored and

  loganhart02

  committed
  5beb0b3

  Copy full SHA for 5beb0b3

• Fix SpeakerManager usage in synthesize.py

  

  erogol

  authored and

  loganhart02

  committed
  6ccbf9f

  Copy full SHA for 6ccbf9f

• Restore last_epoch of the scheduler

  

  erogol

  authored and

  loganhart02

  committed
  a6082ba

  Copy full SHA for a6082ba

• Make optional to detach duration predictor input

  

  erogol

  authored and

  loganhart02

  committed
  3933208

  Copy full SHA for 3933208

• Update docstring format

  

  erogol

  authored and

  loganhart02

  committed
  407bb78

  Copy full SHA for 407bb78

• Update FastPitch config

  

  erogol

  authored and

  loganhart02

  committed
  69d7807

  Copy full SHA for 69d7807

• Update docstring format

  

  erogol

  authored and

  loganhart02

  committed
  d27ef8d

  Copy full SHA for d27ef8d

• Update docstrings

  

  erogol

  authored and

  loganhart02

  committed
  39897f0

  Copy full SHA for 39897f0

• Update FastPitchLoss

  

  erogol

  authored and

  loganhart02

  committed
  85e98d2

  Copy full SHA for 85e98d2

• Don't use align_score for models with duration predictor

  

  erogol

  authored and

  loganhart02

  committed
  d7042b6

  Copy full SHA for d7042b6

• Format style of the recipes

  

  erogol

  authored and

  loganhart02

  committed
  2b06370

  Copy full SHA for 2b06370

• Refactor FastPitch model

  

  erogol

  authored and

  loganhart02

  committed
  d51f84f

  Copy full SHA for d51f84f

• Update FastPitch don't detach duration network inputs

  

  erogol

  authored and

  loganhart02

  committed
  73d0bb8

  Copy full SHA for 73d0bb8

• Enable aligner for FastPitch

  

  erogol

  authored and

  loganhart02

  committed
  302ed58

  Copy full SHA for 302ed58

• Refactor FastPitchv2

  

  erogol

  authored and

  loganhart02

  committed
  ff6725b

  Copy full SHA for ff6725b

• Disable autcast for criterions

  

  erogol

  authored and

  loganhart02

  committed
  f125e0d

  Copy full SHA for f125e0d

• Add sort_by_audio_len option

  

  erogol

  authored and

  loganhart02

  committed
  a8ac1c4

  Copy full SHA for a8ac1c4

• Implement binary alignment loss

  

  erogol

  authored and

  loganhart02

  committed
  9f43989

  Copy full SHA for 9f43989

• Add PitchExtractor and return dict by collate

  

  erogol

  authored and

  loganhart02

  committed
  ec7c77b

  Copy full SHA for ec7c77b

• Add AlignerNetwork

  

  erogol

  authored and

  loganhart02

  committed
  80bf855

  Copy full SHA for 80bf855

• FastPitch refactor and commenting

  

  erogol

  authored and

  loganhart02

  committed
  d8ef0f5

  Copy full SHA for d8ef0f5

• Update generic.FFTransformer

  

  erogol

  authored and

  loganhart02

  committed
  d13d5bc

  Copy full SHA for d13d5bc

• Add tests for certain FastPitch functions

  

  erogol

  authored and

  loganhart02

  committed
  af26a5d

  Copy full SHA for af26a5d

• Update PositionalEncoding

  

  erogol

  authored and

  loganhart02

  committed
  8f7cc47

  Copy full SHA for 8f7cc47

• Integrate Scarf pixel

  

  erogol

  authored and

  loganhart02

  committed
  8e581de

  Copy full SHA for 8e581de

• Update README.md format

  

  erogol

  authored and

  loganhart02

  committed
  8311558

  Copy full SHA for 8311558

• Refactor TTSDataset

  Show description for c46e987
  

  erogol

  authored and

  loganhart02

  committed
  c46e987

  Copy full SHA for c46e987

• Fix attn mask reading bug

  

  erogol

  authored and

  loganhart02

  committed
  00aa649

  Copy full SHA for 00aa649

• Fix loader setup in base_tts

  

  erogol

  authored and

  loganhart02

  committed
  b575e85

  Copy full SHA for b575e85

• Plot unnormalized pitch by FastPitch

  

  erogol

  authored and

  loganhart02

  committed
  181a781

  Copy full SHA for 181a781

• Reformat multi-speaker handling in GlowTTS

  

  erogol

  authored and

  loganhart02

  committed
  eb4717e

  Copy full SHA for eb4717e

• Plot pitch over spectrogram

  

  erogol

  authored and

  loganhart02

  committed
  d0d8fd2

  Copy full SHA for d0d8fd2

• Use pyworld for pitch

  

  erogol

  authored and

  loganhart02

  committed
  09cc932

  Copy full SHA for 09cc932

• Update loader tests for dict return

  

  erogol

  authored and

  loganhart02

  committed
  6c8184b

  Copy full SHA for 6c8184b

• Fix linter issues

  

  erogol

  authored and

  loganhart02

  committed
  6cb6034

  Copy full SHA for 6cb6034

• Add FastPitch model to .models.json

  

  erogol

  authored and

  loganhart02

  committed
  85c3c26

  Copy full SHA for 85c3c26

• Bump up to v0.2.2

  

  erogol

  authored and

  loganhart02

  committed
  1e118c2

  Copy full SHA for 1e118c2

• added command line training file for ljspeech tacotron2 models.

  loganhart02

  committed
  d155e97

  Copy full SHA for d155e97

• added recipe for sc glow tts training on vctk and tacotron2 models for sam dataset, also added options so users can turn on forward and loction attention when they want to.

  loganhart02

  committed
  dc69fc1

  Copy full SHA for dc69fc1

• hifi gan and wavegrad configs for and ljspeech vocoder recipe

  loganhart02

  committed
  97b94ff

  Copy full SHA for 97b94ff

• made ljspeech tts trainer for all models( define models by string instead of calling a function) added more ljspeech vocoder models(these are just baselines though, once I have a free gpu im going …

  Show description for b85160b
  loganhart02

  committed
  b85160b

  Copy full SHA for b85160b

• made ljspeech tts and vocoder trainer.

  loganhart02

  committed
  fd63352

  Copy full SHA for fd63352

• vocoder command line file

  loganhart02

  committed
  2672bb5

  Copy full SHA for 2672bb5

• changed camel case to _. created single speaker and multispeaker auto tts functions

  loganhart02

  committed
  16ff386

  Copy full SHA for 16ff386

• made single speaker vocoder trainer function

  loganhart02

  committed
  ee19679

  Copy full SHA for ee19679

• Merge branch 'recipe_api' of https://github.com/loganhart420/TTS into recipe_api

  loganhart02

  committed
  026249d

  Copy full SHA for 026249d

Commits on Sep 12, 2021

• refactored model configs

  loganhart02

  committed
  7c4dbf7

  Copy full SHA for 7c4dbf7

• refactored model configs

  loganhart02

  committed
  13f91aa

  Copy full SHA for 13f91aa

• refactored model configs

  loganhart02

  committed
  b40e0ba

  Copy full SHA for b40e0ba

• refactored model configs

  loganhart02

  committed
  12fe775

  Copy full SHA for 12fe775

• refactored model configs

  loganhart02

  committed
  a9945d4

  Copy full SHA for a9945d4

• refactored model configs

  loganhart02

  committed
  fdc9220

  Copy full SHA for fdc9220

Commits on Sep 13, 2021

• Merge branch 'coqui-ai:main' into recipe_api

  loganhart02

  authored
  591df9a

  Copy full SHA for 591df9a

Commits on Sep 18, 2021

• added args and usage docs

  loganhart02

  committed
  05e443e

  Copy full SHA for 05e443e

• Merge branch 'coqui-ai:main' into recipe_api

  loganhart02

  authored
  e1e70a0

  Copy full SHA for e1e70a0

Commits on Sep 21, 2021

• documentation for how to train

  loganhart02

  committed
  3062c80

  Copy full SHA for 3062c80

Commits on Oct 18, 2021

• added tacotron2 multispeaker model

  loganhart02

  committed
  49986d2

  Copy full SHA for 49986d2

Commits on Oct 21, 2021

• added pretrained model loading and ljspeech fast pitch model config. added functions to pick forward tts encoder and decoder

  loganhart02

  committed
  77dda0f

  Copy full SHA for 77dda0f

Commits on Oct 26, 2021

• Merge pull request #891 from coqui-ai/dev

  Show description for 40c17b2
  erogol

  authored
  40c17b2

  Copy full SHA for 40c17b2

Commits on Oct 28, 2021

• Merge branch 'main' of https://github.com/coqui-ai/TTS into recipe_api

  Show description for b31d147
  loganhart02

  committed
  b31d147

  Copy full SHA for b31d147

• This is just a class to download some public tts public datasets. still working on adding more this is what I got so far

  loganhart02

  committed
  90cb11a

  Copy full SHA for 90cb11a

• make style

  loganhart02

  committed
  450d450

  Copy full SHA for 450d450

Commits on Oct 29, 2021

• Merge pull request #901 from coqui-ai/dev

  Show description for 33aa27e
  erogol

  authored
  33aa27e

  Copy full SHA for 33aa27e

Commits on Nov 2, 2021

• Merge branch 'coqui-ai:main' into recipe_api

  loganhart02

  authored
  5d5ed65

  Copy full SHA for 5d5ed65