output_format selects the encoding of the generated audio. BreezeBlue generates native 24 kHz mono audio.
Formats by endpoint
The API reference lists selectable format values. HTTP streaming and realtime accept only the encoding names shown above.
Output encodings
AAC and Opus use target bitrates; measured averages vary with the audio. Opus decoder playback rates follow the codec’s rules.
Conversion profiles
A profile usescodec_sample_rate_bitrate, omitting the bitrate for WAV, PCM, and FLAC. Sample rates are in Hz and bitrates in kbps.
Profiles apply to sync and async conversion. Invalid formats return
BAD_REQUEST.
For async jobs, output_format identifies the requested format and audio_format describes the encoding settings.
Realtime output
Chooseoutput_format when creating a realtime session, or in the query string when connecting with an API key. It stays fixed for the connection. pcm delivers raw PCM; mp3 delivers 24 kHz mono MP3 at 128 kbps. Every turn is encoded independently, and its final audio bytes arrive before turn.done. Discard the unplayed remainder of a cancelled turn.
Use formats with
Convert text to speech
Select an output format for sync or async conversion.
Text to speech
Generate audio with HTTP, SDK, and async requests.
Streaming
Receive audio while speech is being generated.
Realtime text to speech
Select PCM or MP3 for a realtime conversation.

