Medix-AI/coqui-tts
Medix-AI
/
coqui-tts
mirror of https://github.com/coqui-ai/TTS.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
4,118 Commits 80 Branches 98 Tags 314 MiB
Commit Graph

330 Commits

Author SHA1 Message Date
Edresson Casanova 4b59f07946 Support the use of speaker embedding as emotion embedding 2022-06-08 09:52:39 -03:00
Edresson Casanova 360b969c23 Fix rebase issues 2022-06-08 09:52:39 -03:00
Edresson Casanova e069985f17 Add speaker and emotion squeezer layers 2022-06-08 09:52:39 -03:00
Edresson Casanova a309edacb4 Remove VITS conditional flow module 2022-06-08 09:52:39 -03:00
Edresson Casanova a1d0088087 Remove VITS End2End loss 2022-06-08 09:52:38 -03:00
Edresson Casanova ae55bdae6c Fix Lint checks 2022-06-08 09:52:38 -03:00
Edresson Casanova fd1036f4ba Add Noise scale predictor 2022-06-08 09:52:38 -03:00
Edresson Casanova d6d8d0e3e1 Fix the VITS GAN loss 2022-06-08 09:52:38 -03:00
Edresson Casanova e07fcc7a8c Add text encoder adversarial loss on the VITS 2022-06-08 09:52:38 -03:00
Edresson Casanova 4e94b46d5e Add end2end VITS loss 2022-06-08 09:52:38 -03:00
Edresson Casanova a822f21b78 Add prosody encoder inference support 2022-06-08 09:52:38 -03:00
Edresson Casanova 010f847929 Add an option to detach the prosody encoder input 2022-06-08 09:52:38 -03:00
Edresson Casanova 2cac18c7b7 Add VAE prosody encoder 2022-06-08 09:52:37 -03:00
Edresson Casanova f774cf0648 Condition the prosody encoder on z_p 2022-06-08 09:52:37 -03:00
Edresson Casanova 512525cc39 Support prosody conditional model on decoder input 2022-06-08 09:52:37 -03:00
Edresson Casanova 02194367d7 Add emotion classifier loss 2022-06-08 09:52:37 -03:00
Edresson Casanova a6c8fea192 Add conditional module 2022-06-08 09:52:37 -03:00
Edresson Casanova bce4a41b9c Fix unit tests 2022-06-08 09:52:37 -03:00
Edresson Casanova 0fb1b200c6 Fix rebase issues 2022-06-08 09:52:37 -03:00
Edresson Casanova 98c2834b17 Disable the reversal prosody encoder speaker loss 2022-06-08 09:52:37 -03:00
Edresson Casanova ac3f98cefb Add text encoder reversal speaker classifier loss 2022-06-08 09:52:37 -03:00
Edresson Casanova a543d71352 Clean up old code 2022-06-08 09:52:36 -03:00
Edresson Casanova 66e3f5388e Add prosody encoder params on config 2022-06-08 09:52:36 -03:00
Edresson Casanova 95409be0bc Add Speech style balancer 2022-06-08 09:52:36 -03:00
Edresson Casanova 050f7707e2 Add reversal classifier loss 2022-06-08 09:52:36 -03:00
Edresson Casanova 44ec2ab387 Add prosody encoder training support 2022-06-08 09:52:36 -03:00
Edresson Casanova 6126e5e588 Add emotion embedding in the encoder 2022-06-08 09:52:36 -03:00
Edresson Casanova 1fdef1c4c9 Add formatter for the Emotional Speech Dataset 2022-06-08 09:52:36 -03:00
Edresson Casanova 61a04a7855 Remove useless encoder weights reload 2022-06-08 09:52:36 -03:00
Edresson Casanova e8c4417f07 Fix Style tests 2022-06-08 09:52:36 -03:00
Edresson Casanova 730befebcc Fix style tests 2022-06-08 09:52:36 -03:00
Edresson Casanova e409f3588b Bug fix in single speaker emotion embedding training 2022-06-08 09:52:36 -03:00
Edresson Casanova 7a0eba517f Add emotion external embeddings training unit test 2022-06-08 09:52:35 -03:00
Edresson Casanova 5a10ef27b3 Add emotion consistency loss 2022-06-08 09:52:35 -03:00
Edresson Casanova bd99548016 Add Emotion Support for the VITS model 2022-06-08 09:52:35 -03:00
Eren Gölge f70e82cd19
Use fsspec and torch for embedding file IO (#1581) 
* Use fsspec and torch for embedding file

* Fixup

* Fix load and save files

* Fix compute embedding script

* Set use_cuda to true if available

* Add dummy speakers.pth file

* Make style

* Change default speakers file extension

Co-authored-by: WeberJulian <julian.weber@hotmail.fr>
 2022-06-01 13:49:42 +02:00
a-froghyar 8be21ec387
Capacitron (#977) 
* new CI config

* initial Capacitron implementation

* delete old unused file

* fix empty formatting changes

* update losses and training script

* fix previous commit

* fix commit

* Add Capacitron test and first round of test fixes

* revert formatter change

* add changes to the synthesizer

* add stepwise gradual lr scheduler and changes to the recipe

* add inference script for dev use

* feat: add posterior inference arguments to synth methods
- added reference wav and text args for posterior inference
- some formatting

* fix: add espeak flag to base_tts and dataset APIs
- use_espeak_phonemes flag was not implemented in those APIs
- espeak is now able to be utilised for phoneme generation
- necessary phonemizer for the Capacitron model

* chore: update training script and style
- training script includes the espeak flag and other hyperparams
- made style

* chore: fix linting

* feat: add Tacotron 2 support

* leftover from dev

* chore:rename parser args

* feat: extract optimizers
- created a separate optimizer class to merge the two optimizers

* chore: revert arbitrary trainer changes

* fmt: revert formatting bug

* formatting again

* formatting fixed

* fix: log func

* fix: update optimizer
- Implemented load_state_dict for continuing training

* fix: clean optimizer init for standard models

* improvement: purge espeak flags and add training scripts

* Delete capacitronT2.py

delete old training script, new one is pushed

* feat: capacitron trainer methods
- extracted capacitron specific training  operations from the trainer into custom
methods in taco1 and taco2 models

* chore: renaming and merging capacitron and gst style args

* fix: bug fixes from the previous commit

* fix: implement state_dict method on CapacitronOptimizer

* fix: call method

* fix: inference naming

* Delete train_capacitron.py

* fix: synthesize

* feat: update tests

* chore: fix style

* Delete capacitron_inference.py

* fix: fix train tts t2 capacitron tests

* fix: double forward in T2 train step

* fix: double forward in T1 train step

* fix: run make style

* fix: remove unused import

* fix: test for T1 capacitron

* fix: make lint

* feat: add blizzard2013 recipes

* make style

* fix: update recipes

* chore: make style

* Plot test sentences in Tacotron

* chore: make style and fix import

* fix: call forward first before problematic floordiv op

* fix: update recipes

* feat: add min_audio_len to recipes

* aux_input["style_mel"]

* chore: make style

* Make capacitron T2 recipe more stable

* Remove T1 capacitron Ljspeech

* feat: implement new grad clipping routine and update configs

* make style

* Add pretrained checkpoints

* Add default vocoder

* Change trainer package

* Fix grad clip issue for tacotron

* Fix scheduler issue with tacotron

Co-authored-by: Eren Gölge <egolge@coqui.ai>
Co-authored-by: WeberJulian <julian.weber@hotmail.fr>
Co-authored-by: Eren Gölge <erogol@hotmail.com>
 2022-05-20 16:17:11 +02:00
Edresson Casanova ee99a6c1e2 Fix voice conversion inference (#1583) 
* Add voice conversion zoo test

* Fix style

* Fix unit test
 2022-05-20 15:50:25 +02:00
Edresson Casanova e5d8ec2402
Change the VITS upsampling interpolation trick to linear (#1564) 2022-05-13 10:52:39 +02:00
Edresson Casanova c6008e5235
Add audio length sampler balancer (#1561) 
* Add audio length sampler balancer

* Add unit tests
 2022-05-12 19:59:19 +02:00
Eren Gölge 6e460b7e42
Add an assert for the upsampling trick (#1538) 2022-05-12 19:55:24 +02:00
Eren Gölge e45ae57aef
Merge pull request #1550 from coqui-ai/fix-upsampling-asserts 
Fix VITS upsampling asserts
 2022-05-12 14:51:41 +02:00
Edresson Casanova 175ca06388 Add reinit text encoder and duration predictor parameter (#1562) 
* Add reinit encoder and duration predictor option

* Add .data to prevent any overlooked autograd hook
 2022-05-12 09:08:36 -03:00
Edresson Casanova 182711043c Fix the VITS upsampling asserts 
Fix style
 2022-05-12 09:08:29 -03:00
Eren Gölge c18bd21b3f Return durations at VITS inference 2022-05-11 11:30:05 +02:00
Eren Gölge 5021a03de0 Use torch.no_grad for VITS inference 2022-05-11 11:29:36 +02:00
Eren Gölge 3f03e3012c Fix batch_group_size in VITS 2022-05-07 13:44:44 +02:00
WeberJulian fbdf76b2fc returns y_mask in VITS inference (#1540) 
* returns y_mask

* make style
 2022-05-03 13:49:24 +02:00
Edresson Casanova 8d228ab22a
Trick to Upsampling to High sampling rates using VITS model (#1456) 
* Add upsample VITS support

* Fix the bug in inference

* Fix lint checks

* Add RMS based norm in save_wav method

* Style fix

* Add the period for VITS multi-period discriminator in model_args

* Bug fix in speaker encoder load in inference time

* Add unit tests

* Remove useless detach_z_vocoder parameter

* Add docs for VITS upsampling

* Fix the docs

* Rename TTS_part_sample_rate to encoder_sample_rate

* Add upsampling_init and upsampling_z methods

* Add asserts for encoder_sample_rate part

* Move upsampling tests to test_vits.py
 2022-04-26 11:47:46 +02:00
Edresson Casanova 060e0f9368
Add EmbeddingManager and BaseIDManager (#1374) 2022-03-31 13:41:16 +02:00