diff --git a/examples/canton/tts3/README.md b/examples/canton/tts3/README.md
index 87ef40907..d2c46f642 100644
--- a/examples/canton/tts3/README.md
+++ b/examples/canton/tts3/README.md
@@ -37,6 +37,7 @@ Run the command below to
 3. train the model.
 4. synthesize wavs.
     - synthesize waveform from `metadata.jsonl`.
+    - `--stage` controls the vocoder model during synthesis (0 = pwgan, 1 = hifigan).
     - synthesize waveform from text file.
 ```bash
 ./run.sh
diff --git a/examples/canton/tts3/run.sh b/examples/canton/tts3/run.sh
index acfc50223..29502a964 100755
--- a/examples/canton/tts3/run.sh
+++ b/examples/canton/tts3/run.sh
@@ -29,12 +29,12 @@ 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} ${ckpt_name} || exit -1
+    CUDA_VISIBLE_DEVICES=${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} ${ckpt_name} || exit -1
+    CUDA_VISIBLE_DEVICES=${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