【PaddleSpeech No.7】 (#4043)

Echo-Nie · web-flow · commit de3fa28b1bde · 2025-04-07T15:12:13.000+08:00
* run.sh&#20462;&#25913;&#65306;&#20026; synthesize &#21644; synthesize_e2e &#28155;&#21152; --stage &#21442;&#25968;&#25511;&#21046; vocoder &#27169;&#22411;&#36873;&#25321;&#65292;REAMDE.md&#20462;&#25913;&#65306;&#34917;&#20805; stage &#21442;&#25968;&#35828;&#26126;&#65292;&#26126;&#30830; vocoder &#36873;&#25321;&#36923;&#36753;

* &#28155;&#21152;run.sh&#20013;stage&#21442;&#25968;&#30456;&#20851;&#30340;&#27880;&#37322;

* HiFiGAN&#25913;&#20026;MultiBand MelGAN

* cmsc&#25991;&#20214;&#25913;&#22238;&#21407;&#20301;&#65288;No.15&#19981;&#20462;&#25913;&#65289;&#65292;&#36825;&#37324;&#21482;&#23545;No.6&#20570;&#20462;&#25913;

* fix the tts0
diff --git a/examples/csmsc/tts0/README.md b/examples/csmsc/tts0/README.md
@@ -99,8 +99,10 @@ pwg_baker_ckpt_0.4
 ```
 `./local/synthesize.sh` calls `${BIN_DIR}/../synthesize.py`, which can synthesize waveform from `metadata.jsonl`.
 ```bash
-CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh ${conf_path} ${train_output_path} ${ckpt_name}
+CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name}
 ```
+`--stage` controls the vocoder model during synthesis, which can use stage `0-4` to select the vocoder to use {`pwgan`, `multi band melgan`, `style melgan`, ` hifigan`, `wavernn`}
+
 ```text
 usage: synthesize.py [-h]
                      [--am {speedyspeech_csmsc,fastspeech2_csmsc,fastspeech2_ljspeech,fastspeech2_aishell3,fastspeech2_vctk,tacotron2_csmsc,tacotron2_ljspeech,tacotron2_aishell3}]
@@ -146,9 +148,12 @@ optional arguments:
                         output dir.
 ```
 `./local/synthesize_e2e.sh` calls `${BIN_DIR}/../synthesize_e2e.py`, which can synthesize waveform from text file.
+
 ```bash
-CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh ${conf_path} ${train_output_path} ${ckpt_name}
+CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name}
 ```
+`--stage` controls the vocoder model during synthesis, which can use stage `0,1,3,4` to select the vocoder to use{`pwgan`, `multi band melgan`, `hifigan`, `wavernn`}
+
 ```text
 usage: synthesize_e2e.py [-h]
                          [--am {speedyspeech_csmsc,speedyspeech_aishell3,fastspeech2_csmsc,fastspeech2_ljspeech,fastspeech2_aishell3,fastspeech2_vctk,tacotron2_csmsc,tacotron2_ljspeech}]
diff --git a/examples/csmsc/tts0/run.sh b/examples/csmsc/tts0/run.sh
@@ -27,13 +27,15 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
 fi
 
 if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
-    # synthesize, vocoder is pwgan
-    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
+    # synthesize, vocoder is pwgan by default stage 0
+    # stage 1-4 to select the vocoder to use {multi band melgan, style melgan, hifigan, wavernn}
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
 fi
 
 if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
-    # synthesize_e2e, vocoder is pwgan
-    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
+    # synthesize_e2e, vocoder is pwgan by default stage 0
+    # stage 1,3,4 to select the vocoder to use {multi band melgan, hifigan, wavernn}
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
 fi
 
 if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
