add svc inference code

example/infer_svc.sh

@@ -0,0 +1,27 @@
+#!/bin/bash
+
+script_dir=$(dirname "$(realpath "$0")")
+root_dir=$(dirname "$script_dir")
+
+cd $root_dir || exit
+export PYTHONPATH=$root_dir:$PYTHONPATH
+
+model_path=pretrained_models/SoulX-Singer-SVC/model.pt
+config=soulxsinger/config/soulxsinger.yaml
+prompt_wav_path=example/audio/zh_prompt.mp3
+target_wav_path=example/audio/music.mp3
+prompt_f0_path=example/audio/zh_prompt_f0.npy
+target_f0_path=example/audio/music_f0.npy
+save_dir=example/generated/music_svc
+
+python -m cli.inference_svc \
+    --device cuda \
+    --model_path $model_path \
+    --config $config \
+    --prompt_wav_path $prompt_wav_path \
+    --target_wav_path $target_wav_path \
+    --prompt_f0_path $prompt_f0_path \
+    --target_f0_path $target_f0_path \
+    --save_dir $save_dir \
+    --auto_shift \
+    --pitch_shift 0

example/preprocess.sh

@@ -15,6 +15,7 @@ save_dir=example/transcriptions/zh_prompt
 language=Mandarin
 vocal_sep=False
 max_merge_duration=30000
+midi_transcribe=True    # Whether to transcribe vocal midi, set True for singing voice synthesis, False for singing voice conversion
 
 python -m preprocess.pipeline \
     --audio_path $audio_path \
@@ -22,7 +23,8 @@ python -m preprocess.pipeline \
     --language $language \
     --device $device \
     --vocal_sep $vocal_sep \
-    --max_merge_duration $max_merge_duration
+    --max_merge_duration $max_merge_duration \
+    --midi_transcribe $midi_transcribe
 
 
 ####### Run Target Annotation #######
@@ -31,6 +33,7 @@ save_dir=example/transcriptions/music
 language=Mandarin
 vocal_sep=True
 max_merge_duration=60000
+midi_transcribe=True    # Whether to transcribe vocal midi, set True for singing voice synthesis, False for singing voice conversion
 
 python -m preprocess.pipeline \
     --audio_path $audio_path \
@@ -38,4 +41,5 @@ python -m preprocess.pipeline \
     --language $language \
     --device $device \
     --vocal_sep $vocal_sep \
-    --max_merge_duration $max_merge_duration
+    --max_merge_duration $max_merge_duration \
+    --midi_transcribe $midi_transcribe