Respond in JSON format to divide the different captions in frontend

When OpenAI includes the possibility to use the respond format this will be implemented