Home Audio Decision API directly answers audio questions Home has released the Audio Decision API. It requires a tone and a question and returns a structured response with a confidence score. Examples for developers: determine the number of speakers, detect laughter or check if interlocutors disturb each other. The usual scheme first converts speech to text and then sends the transcript to the language model. In this case, some information about pauses, intonation and sounds will be lost. The Home API analyzes the source audio and is suitable as a decision-making layer for the voice agent: recognition and playback can remain separate components. In the demo recording, Home shows batch call processing. For long files, the documentation suggests splitting the record into chunks and running queries on the chunks. The confidence value should be checked against your data: it does not guarantee the correct interpretation of emotions. You can request API access on the product page. The voice of Mekorot.

Home Audio Decision API directly answers audio questions Home has released the Audio Decision API. It requires a tone and a question and returns a structured response with a confidence score. Examples for developers: determine the number of speakers, detect laughter or check if interlocutors disturb each other. The usual scheme first converts speech to text and then sends the transcript to the language model. In this case, some information about pauses, intonation and sounds will be lost. The Home API analyzes the source audio and is suitable as a decision-making layer for the voice agent: recognition and playback can remain separate components. In the demo recording, Home shows batch call processing. For long files, the documentation suggests splitting the record into chunks and running queries on the chunks. The confidence value should be checked against your data: it does not guarantee the correct interpretation of emotions. You can request API access on the product page. The voice of Mekorot.