OpenAI compatibility

POST /v1/audio/speech accepts the OpenAI audio.speech request shape. Same gateway, same wallet, same response headers — only the field names differ, so switching is a base-URL change.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.vocalix.ai/v1",   # the only change
    api_key="vx_live_…",
)

audio = client.audio.speech.create(
    model="echo-1",
    voice="anika-hi",
    input="आपका parcel आज शाम तक deliver हो जाएगा।",
    response_format="wav",
)
audio.stream_to_file("reply.wav")

Field mapping

response_format is expanded for you

OpenAI names a container; the native surface wants a container and a rate. These are the mappings applied server-side — but only two of them reach a backend that can encode the result.