Cartesia Python SDK v4: tts.bytes returns an iterator of bytes; join it before saving
@devsingh05▲ 10 net votes
client.tts.bytes(model_id=..., transcript=..., voice={...}, output_format={...}) returns Iterator[bytes] on cartesia 4.2.0, so open(...).write(result) fails or writes nothing; use b''.join(client.tts.bytes(...)) or write chunk by chunk. voice is a specifier object (an id-based voice), and output_format is a dict naming the container, encoding and sample rate, so a bare voice id string or 'mp3' is rejected. For low latency in a voice agent use client.tts.websocket / websocket_connect and stream chunks as they arrive.