OpenAI Whisper Configuration
caution
For openai-whisper, the language parameter is optional. However, if you choose to set it, you need to use the correct language code. The language codes for openai-whisper are different from those used in azure-stt. For example, for Chinese in openai-whisper, you would use the code zh. In the roles.json file, you can configure different language codes for each role. You can find a list of all language codes for openai-whisper here: OpenAI Whisper language 639-1 codes
Configurable parameters:
| Parameter Name | Description | Default Value |
|---|---|---|
| language | The language of the input audio. Supplying the input language in ISO-639-1 format will improve accuracy and latency. | None |
| api_base | The base URL of the API endpoint. | https://api.openai.com/v1 |
| key | Your API key. | None |
| model | The ID of the model to use. | whisper-1 |
| prompt | An optional text to guide the model's style or continue a previous audio segment. The prompt should match the audio language. | "ok. Let's begin." |
Configuration example:
roles.json
{
"1": {
"start_text": "你好,我是小兔兔,请问有什么我可以帮助你的吗?",
"prompt": "你扮演一个孩子的小伙伴,名字叫小兔兔,性格和善,说话活泼可爱,对孩子充满爱心,经常赞赏和鼓励孩子,用5岁孩子容易理解语言提供有趣和创新的回答,每次回复根据聊天主题询问她的看法以激发她的思考和好奇心,现在她来到了你身边问了第一个问题:[你是谁]",
"stt_type": "openai-whisper",
"stt_config": {
"language": "zh",
"api_base": "https://api.openai.com/v1",
"key": "sk-AAAAAAAAAAAAAAAAAAa",
"model": "whisper-1",
"prompt": "ok. Let's begin."
}
}
}