diff --git a/examples/csmsc/tts3_rhy/README.md b/examples/csmsc/tts3_rhy/README.md
index 855aa885c..da6191592 100644
--- a/examples/csmsc/tts3_rhy/README.md
+++ b/examples/csmsc/tts3_rhy/README.md
@@ -14,11 +14,13 @@ Remember in our repo, you should add `--rhy-with-duration` flag to obtain the rh
 Assume the path to the dataset is `~/datasets/BZNSYP`.
 Assume the path to the MFA result of CSMSC is `./baker_alignment_tone`.
 Run the command below to
+
 1. **source path**.
 2. preprocess the dataset.
 3. train the model.
 4. synthesize wavs.
     - synthesize waveform from `metadata.jsonl`.
+    - select vocoder type via `--stage` (0 = pwgan, 1 = hifigan)
     - synthesize waveform from a text file.
 5. inference using the static model.
 ```bash
diff --git a/examples/csmsc/tts3_rhy/run.sh b/examples/csmsc/tts3_rhy/run.sh
index e49f43ee6..61020093b 100755
--- a/examples/csmsc/tts3_rhy/run.sh
+++ b/examples/csmsc/tts3_rhy/run.sh
@@ -28,11 +28,11 @@ if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
 fi
 
 if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
-    # synthesize, vocoder is pwgan by default
-    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
+    # synthesize, vocoder is pwgan by default stage 0, stage 1 will use hifigan as vocoder
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
 fi
 
 if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
-    # synthesize_e2e, vocoder is pwgan by default
-    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
+    # synthesize_e2e, vocoder is pwgan by default stage 0, stage 1 will use hifigan as vocoder
+    CUDA_VISIBLE_DEVICES=${gpus} ./local/synthesize_e2e.sh --stage 0 ${conf_path} ${train_output_path} ${ckpt_name} || exit -1
 fi