AzureOpenAI.audio.speech.create(...) rejects instructions parameter

[adarshxs]

Confirm this is an issue with the Python library and not an underlying OpenAI API

• This is an issue with the Python library

Describe the bug

When using the AzureOpenAI client from the openai Python package with the audio.speech.create(...) or audio.speech.with_streaming_response.create(...) methods, passing an instructions argument results in a Python TypeError. However, the official OpenAI API docs show instructions as a supported parameter for text-to-speech.

when I send a curl request with instructions in data field, it works as well, so I'm guessing its not an endpoint issue but rather AzureOpenAI client?

To Reproduce

1. Run the code snipper below

from openai import AzureOpenAI
import os

client = AzureOpenAI(
    azure_endpoint=os.getenv("AZURE_OPENAI_ENDPOINT"),
    api_key=os.getenv("AZURE_OPENAI_API_KEY"),
    api_version="2025‑03‑01‑preview"
)

with client.audio.speech.with_streaming_response.create(
    model="gpt‑4o‑mini‑tts",
    voice="coral",
    input="Today is a wonderful day!",
    instructions="Speak in a cheerful and positive tone."
) as resp:
    resp.stream_to_file("out.mp3")

Output:
TypeError: Speech.create() got an unexpected keyword argument 'instructions'

Code snippets

This works well:


curl -X POST "https://xxxx.cognitiveservices.azure.com/openai/deployments/tts/audio/speech..." \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -d '{
      "model": "tts",
      "input": "The quick brown fox jumped over the lazy dog",
      "voice": "alloy",
      "instructions": "scary and dark and deep voice",
      "response_format": "wav"
    }' --output output.wav

OS

Windows

Python version

python v3.11.7

Library version

openai v1.88.0

[Vishal-sys-code]

Hi, I've investigated this.

It seems the core library files (SpeechCreateParams in types/audio/speech_create_params.py and the Speech.create method in resources/audio/speech.py) in v1.88.0 correctly include the instructions parameter. The standard inheritance path for AzureOpenAI doesn't show an obvious override that would remove it.

The TypeError when using AzureOpenAI.audio.speech.create(instructions=...), despite the parameter being defined in the base classes and the cURL working, is puzzling.

This might point to an issue where the effective runtime method signature for AzureOpenAI's audio.speech.create differs from the base version, possibly due to the code generation process (Stainless) for Azure clients, or a very specific environment interaction.

Could this be an Azure-specific generation difference? It would be great if someone familiar with the Azure client generation could weigh in.

[adarshxs]

cc @RobertCraigie @kristapratico