Create speech

Start with the required fields and an output format.

speech = client.audio.speech.create(
    model="teen-v1",
    input="Create a Teen V1 response.",
    response_format="mp3",
)
speech.write_to_file("speech.mp3")

Required fields

Field	Value
`model`	`teen-v1`
`input`	1–4096 characters

voice is optional and defaults to default. Supply a public or saved voice_... ID only when selecting that voice.

Output fields

Field	Default	Accepted values
`response_format`	`pcm`	`pcm`, `mp3`, `wav`
`output_sample_rate`	Server default	8000–48000 Hz
`stream`	`True`	`True` or `False`; WAV requires `False`
`stream_format`	`audio`	`audio` or `sse`

Generation fields

Field	Default	Purpose
`quality`	`fast`	Use `expressive` for the expressive generation profile
`cfg_value`	`2.0`	Guidance strength from 1.0 to 3.0
`max_generate_length`	`2000`	Model generation budget from 32 to 4096
`seed`	None	Improve repeatability
`lora`	None	Select a registered server-side LoRA

Reference audio, reference transcripts, and request-time voice descriptions are covered in voice modelling.