diff --git a/examples/fastspeech/README.md b/examples/fastspeech/README.md index 865f68ae0af04e935360eea51a704ad960626500..0c40488db4c4acf5dbc6def70aa9a5a7322fcbc3 100644 --- a/examples/fastspeech/README.md +++ b/examples/fastspeech/README.md @@ -126,10 +126,10 @@ python synthesis.py \ --checkpoint=${CHECKPOINTPATH} \ --config='configs/ljspeech.yaml' \ --output=${OUTPUTPATH} \ ---vocoder='griffinlim' \ +--vocoder='griffin-lim' \ ``` -We currently support two vocoders, ``griffinlim`` and ``waveflow``. You can set ``--vocoder`` to use one of them. If you want to use ``waveflow`` as your vocoder, you need to set ``--config_vocoder`` and ``--checkpoint_vocoder`` which are the path of the config and checkpoint of vocoder. You can download the pretrain model of ``waveflow`` from [here](https://github.com/PaddlePaddle/Parakeet#vocoders). +We currently support two vocoders, ``Griffin-Lim`` algorithm and ``WaveFlow``. You can set ``--vocoder`` to use one of them. If you want to use ``waveflow`` as your vocoder, you need to set ``--config_vocoder`` and ``--checkpoint_vocoder`` which are the path of the config and checkpoint of vocoder. You can download the pre-trained model of ``waveflow`` from [here](https://github.com/PaddlePaddle/Parakeet#vocoders). Or you can run the script file directly. diff --git a/examples/fastspeech/synthesis.py b/examples/fastspeech/synthesis.py index 96eceb5f062e7f06bd8db730774ded4d59a1bbfb..dde776ff7763aa69a7d29b4ab5017a3a234d37ce 100644 --- a/examples/fastspeech/synthesis.py +++ b/examples/fastspeech/synthesis.py @@ -39,8 +39,8 @@ def add_config_options_to_parser(parser): parser.add_argument( "--vocoder", type=str, - default="griffinlim", - choices=['griffinlim', 'waveflow'], + default="griffin-lim", + choices=['griffin-lim', 'waveflow'], help="vocoder method") parser.add_argument( "--config_vocoder", type=str, help="path of the vocoder config file") @@ -53,11 +53,11 @@ def add_config_options_to_parser(parser): ) parser.add_argument( - "--checkpoint", type=str, help="fastspeech checkpoint to synthesis") + "--checkpoint", type=str, help="fastspeech checkpoint for synthesis") parser.add_argument( "--checkpoint_vocoder", type=str, - help="vocoder checkpoint to synthesis") + help="vocoder checkpoint for synthesis") parser.add_argument( "--output", @@ -96,7 +96,7 @@ def synthesis(text_input, args): _, mel_output_postnet = model(text, pos_text, alpha=args.alpha) - if args.vocoder == 'griffinlim': + if args.vocoder == 'griffin-lim': #synthesis use griffin-lim wav = synthesis_with_griffinlim(mel_output_postnet, cfg['audio']) elif args.vocoder == 'waveflow': diff --git a/examples/fastspeech/synthesis.sh b/examples/fastspeech/synthesis.sh index a94376f66b3cc345f8470bd3bf80f3f3251f4c5a..1ebed1bf57017969913a55e7c502731161a0d9c1 100644 --- a/examples/fastspeech/synthesis.sh +++ b/examples/fastspeech/synthesis.sh @@ -4,12 +4,12 @@ CUDA_VISIBLE_DEVICES=0 \ python -u synthesis.py \ --use_gpu=1 \ --alpha=1.0 \ ---checkpoint='./checkpoint/fastspeech/step-162000' \ ---config='configs/ljspeech.yaml' \ +--checkpoint='./fastspeech_ljspeech_ckpt_1.0/fastspeech/step-162000' \ +--config='fastspeech_ljspeech_ckpt_1.0/ljspeech.yaml' \ --output='./synthesis' \ --vocoder='waveflow' \ ---config_vocoder='../waveflow/checkpoint/waveflow_res128_ljspeech_ckpt_1.0/waveflow_ljspeech.yaml' \ ---checkpoint_vocoder='../waveflow/checkpoint/waveflow_res128_ljspeech_ckpt_1.0/step-2000000' \ +--config_vocoder='./waveflow_res128_ljspeech_ckpt_1.0/waveflow_ljspeech.yaml' \ +--checkpoint_vocoder='./waveflow_res128_ljspeech_ckpt_1.0/step-2000000' \ diff --git a/examples/transformer_tts/README.md b/examples/transformer_tts/README.md index b449c6ab6e8e5deb710425eee782bc6ba8b329b6..e8e013186f09e3608c8d397bc651ecc16eb8f5e9 100644 --- a/examples/transformer_tts/README.md +++ b/examples/transformer_tts/README.md @@ -95,10 +95,10 @@ python synthesis.py \ --output=${OUTPUTPATH} \ --config='configs/ljspeech.yaml' \ --checkpoint_transformer=${CHECKPOINTPATH} \ ---vocoder='griffinlim' \ +--vocoder='griffin-lim' \ ``` -We currently support two vocoders, ``griffinlim`` and ``waveflow``. You can set ``--vocoder`` to use one of them. If you want to use ``waveflow`` as your vocoder, you need to set ``--config_vocoder`` and ``--checkpoint_vocoder`` which are the path of the config and checkpoint of vocoder. You can download the pretrain model of ``waveflow`` from [here](https://github.com/PaddlePaddle/Parakeet#vocoders). +We currently support two vocoders, ``Griffin-Lim`` algorithm and ``WaveFlow``. You can set ``--vocoder`` to use one of them. If you want to use ``waveflow`` as your vocoder, you need to set ``--config_vocoder`` and ``--checkpoint_vocoder`` which are the path of the config and checkpoint of vocoder. You can download the pre-trained model of ``waveflow`` from [here](https://github.com/PaddlePaddle/Parakeet#vocoders). Or you can run the script file directly. diff --git a/examples/transformer_tts/synthesis.py b/examples/transformer_tts/synthesis.py index 946463824f6bc354a8410ec8c373094251d5e8dc..effbffdc7a0cc21793a72eecd3ba39ead469ecce 100644 --- a/examples/transformer_tts/synthesis.py +++ b/examples/transformer_tts/synthesis.py @@ -49,19 +49,19 @@ def add_config_options_to_parser(parser): parser.add_argument( "--checkpoint_transformer", type=str, - help="transformer_tts checkpoint to synthesis") + help="transformer_tts checkpoint for synthesis") parser.add_argument( "--vocoder", type=str, - default="griffinlim", - choices=['griffinlim', 'waveflow'], + default="griffin-lim", + choices=['griffin-lim', 'waveflow'], help="vocoder method") parser.add_argument( "--config_vocoder", type=str, help="path of the vocoder config file") parser.add_argument( "--checkpoint_vocoder", type=str, - help="vocoder checkpoint to synthesis") + help="vocoder checkpoint for synthesis") parser.add_argument( "--output", @@ -124,7 +124,7 @@ def synthesis(text_input, args): i * 4 + j, dataformats="HWC") - if args.vocoder == 'griffinlim': + if args.vocoder == 'griffin-lim': #synthesis use griffin-lim wav = synthesis_with_griffinlim(postnet_pred, cfg['audio']) elif args.vocoder == 'waveflow': diff --git a/examples/transformer_tts/synthesis.sh b/examples/transformer_tts/synthesis.sh index 1ceee837e0b5bd1db568de0ba7ff2b4348b53c8b..be91cd4441c6c0e0d63dedf547aa108551238e69 100644 --- a/examples/transformer_tts/synthesis.sh +++ b/examples/transformer_tts/synthesis.sh @@ -4,11 +4,11 @@ CUDA_VISIBLE_DEVICES=0 \ python -u synthesis.py \ --use_gpu=0 \ --output='./synthesis' \ ---config='configs/ljspeech.yaml' \ ---checkpoint_transformer='./checkpoint/transformer/step-120000' \ +--config='transformer_tts_ljspeech_ckpt_1.0/ljspeech.yaml' \ +--checkpoint_transformer='./transformer_tts_ljspeech_ckpt_1.0/step-120000' \ --vocoder='waveflow' \ ---config_vocoder='../waveflow/checkpoint/waveflow_res128_ljspeech_ckpt_1.0/waveflow_ljspeech.yaml' \ ---checkpoint_vocoder='../waveflow/checkpoint/waveflow_res128_ljspeech_ckpt_1.0/step-2000000' \ +--config_vocoder='./waveflow_res128_ljspeech_ckpt_1.0/waveflow_ljspeech.yaml' \ +--checkpoint_vocoder='./waveflow_res128_ljspeech_ckpt_1.0/step-2000000' \ if [ $? -ne 0 ]; then echo "Failed in training!"