When TTS, user input Chinese characters, then the backend parse these characters as None

[Kang-Jack]

===env windows docker desktop============
docker image:
ghcr.io/speaches-ai/speaches:latest-cuda

=========issue===============

with same mode , i can get expected .mp3 file and result when my inputs only contains number characters , but get a empty .mp3 file when i use Chinese characters as input from log looks the input has be erased due to uncorrected json decode. the request and log shows below:

===============correct result when only numbers input=========================================
curl --request POST
--url http://127.0.0.1:8030/v1/audio/speech
--header 'Content-Type: application/json'
--data '{
"input": "1234567",
"model": "speaches-ai/piper-zh_CN-huayan-medium",
"voice": "huayan",
"sample_rate": 8000,
"speed":0.8,
"response_format": "mp3"
}'

log:

2025-11-04 14:53:24 2025-11-04 06:53:24,647:INFO:speaches.model_manager:_decrement_ref:64:Model speaches-ai/piper-zh_CN-huayan-medium is idle, scheduling offload in 300s
2025-11-04 14:53:24 INFO: 172.17.0.1:53484 - "POST /v1/audio/speech HTTP/1.1" 200 OK
2025-11-04 14:53:24 2025-11-04 06:53:24,907:INFO:speaches.executors.piper.utils:generate_audio:154:Generated audio for 7 characters in 0.25853770100002293s

========================empty result when use Chinese characters=============================
curl --request POST
--url http://127.0.0.1:8030/v1/audio/speech
--header 'Content-Type: application/json'
--data '{
"input": "你好，这是关于文字转语音的测试",
"model": "speaches-ai/piper-zh_CN-huayan-medium",
"voice": "huayan",
"sample_rate": 8000,
"speed":0.8,
"response_format": "mp3"
}
'
log:

2025-11-04 14:52:01 2025-11-04 06:52:01,145:DEBUG:speaches.executors.piper.model_manager:load_model:58:speaches-ai/piper-zh_CN-huayan-medium model already loaded
2025-11-04 14:52:01 2025-11-04 06:52:01,145:DEBUG:speaches.model_manager:_increment_ref:54:Model was set to expire in 300s, cancelling
2025-11-04 14:52:01 2025-11-04 06:52:01,145:DEBUG:speaches.model_manager:_increment_ref:56:Incremented ref count for speaches-ai/piper-zh_CN-huayan-medium, self.ref_count=1
2025-11-04 14:52:01 2025-11-04 06:52:01,145:DEBUG:speaches.model_manager:_decrement_ref:61:Decremented ref count for speaches-ai/piper-zh_CN-huayan-medium, self.ref_count=0
2025-11-04 14:52:01 2025-11-04 06:52:01,145:INFO:speaches.model_manager:_decrement_ref:64:Model speaches-ai/piper-zh_CN-huayan-medium is idle, scheduling offload in 300s
2025-11-04 14:52:01 INFO: 172.17.0.1:53182 - "POST /v1/audio/speech HTTP/1.1" 200 OK
2025-11-04 14:52:01 2025-11-04 06:52:01,162:INFO:speaches.executors.piper.utils:generate_audio:154:Generated audio for 0 characters in 0.01593830000001617s