通过apizt实际生成8秒双角色有声视频，固定左侧眼镜角色问“车票买好了吗？”，右侧橙衣角色回答“买好了，明早出发”。ASR把回答识别为疑似错词，需回听并逐帧查看说话者嘴部，不能将有音轨等同于口型已合格。
two-speaker-audio
asr.json
{
"task": "transcribe",
"language": "chinese",
"duration": 8.09000015258789,
"text": "车票买好了吗? 明好明早出发!",
"words": [
{
"word": "车",
"start": 0.800000011920929,
"end": 1.159999966621399
},
{
"word": "票",
"start": 1.159999966621399,
"end": 1.2799999713897705
},
{
"word": "买",
"start": 1.2799999713897705,
"end": 1.5199999809265137
},
{
"word": "好了",
"start": 1.5199999809265137,
"end": 1.7599999904632568
},
{
"word": "吗",
"start": 1.7599999904632568,
"end": 2.0
},
{
"word": "明",
"start": 6.179999828338623,
"end": 6.320000171661377
},
{
"word": "好",
"start": 6.320000171661377,
"end": 6.440000057220459
},
{
"word": "明",
"start": 6.440000057220459,
"end": 6.579999923706055
},
{
"word": "早",
"start": 6.579999923706055,
"end": 6.679999828338623
},
{
"word": "出",
"start": 6.679999828338623,
"end": 6.860000133514404
},
{
"word": "发",
"start": 6.860000133514404,
"end": 7.0
}
],
"usage": {
"type": "duration",
"seconds": 9
}
}