SpeechToSpeechInterfaceConvertSpeechMessage
HTTP API schema
← HTTP API referenceView source specification ↗
Converts the voice in input audio to another voice, keeping the words and delivery. Both audio inputs carry base64 data, at most 5.5 MiB decoded each so that two of them fit the 16 MiB request body, and inputAudio at most 60 seconds; url inputs are rejected (a URL input would need host rules for caller-supplied addresses and is a follow-up).
SpeechToSpeechInterfaceConvertSpeechMessage fields
- model
stringRequired Model identifier from the model catalog (query-models). An unknown model is rejected.
- inputAudioMediaContentRequired
inputAudio: Reference MediaContent
- speakerReferenceMediaContent
speakerReference: Reference MediaContent
- voice
string A preset target voice from the model catalog entry, used when no speakerReference is given
- responseFormatBinaryResponseFormat
responseFormat: Reference BinaryResponseFormat