Discrepancies between Azure Playground / OpenAI native API and Azure API

Question

Discrepancies between Azure Playground / OpenAI native API and Azure API

Louis G 0

I'm facing a bit of a conundrum. I've been using the gpt-4o-mini-tts TTS model through the playground. I'm now trying to transitioning to API calls. In order to direct the TTS speech, I've been using the "instructions" parameter that is exposed through the Azure playground and the native OpenAI API, but I get a HTTP error 400 when using the API due to the inclusion of the "instructions". I seem to have hit an issue where the Azure and OpenAI API surface are out of sync. couple of questions:

Why expose this in the playground if there is no way to use it in the API?
When will this be added to the Azure API?

Lou

Louis G 0 Reputation points

2025-10-08T17:49:20.9966667+00:00

This proposed answer did not answer my issue. Please see my updated below

1 answer

Your answer

Louis G 0 Reputation points

2025-10-08T17:49:20.9966667+00:00

This proposed answer did not answer my issue. Please see my updated below

Answer 1

Divyesh Govaerdhanan 10,310

Hello,

Welcome to Microsoft Q&A,

Instructions are not a valid field on Azure’s /chat/completions request body, so you get HTTP 400. Put “how the voice should speak” in a system message there.

If you want to use a top-level instructions field, call the Responses API instead (that surface supports instructions).

Option A — Chat Completions (no instructions field)

curl -X POST "https://{RESOURCE}.openai.azure.com/openai/deployments/{DEPLOYMENT}/chat/completions?api-version=2025-01-01-preview" \
  -H "api-key: $AZURE_OPENAI_API_KEY" -H "Content-Type: application/json" \
  -d '{
    "modalities": ["text","audio"],
    "audio": { "voice": "alloy", "format": "wav" },
    "messages": [
      { "role": "system", "content": "Speak like a calm British announcer. Keep sentences short." },
      { "role": "user", "content": "Your order is ready for pickup." }
    ]
  }'

This is the preview audio generation flow for gpt-4o(-mini)-audio-preview; the style guidance lives in the system message

https://learn.microsoft.com/en-us/azure/ai-foundry/openai/audio-completions-quickstart

Option B — Responses API (use instructions)

curl -X POST "https://{RESOURCE}.openai.azure.com/openai/v1/responses" \
  -H "api-key: $AZURE_OPENAI_API_KEY" -H "Content-Type: application/json" \
  -d '{
    "model": "{DEPLOYMENT}",                // your gpt-4o / gpt-4o-mini audio-capable deployment
    "instructions": "Speak like a calm British announcer. Keep sentences short.",
    "input": "Your order is ready for pickup.",
    "modalities": ["text","audio"],
    "audio": { "voice": "alloy", "format": "wav" }
  }'

Azure’s Responses API supports instructions (top-level) and is available in multiple regions.

Please Upvote and accept the answer if it helps!!

Louis G 0

My question is in regards to the gpt-4o-mini-tts model specifically. OpenAI's API supports "instructions" where I don't see a way to do this in Azure. Your examples assume I use a different model entirely. https://platform.openai.com/docs/guides/text-to-speech/quickstart

from pathlib import Path
from openai import OpenAI

client = OpenAI()
speech_file_path = Path(__file__).parent / "speech.mp3"

with client.audio.speech.with_streaming_response.create(
    model="gpt-4o-mini-tts",
    voice="coral",
    input="Today is a wonderful day to build something people love!",
    instructions="Speak in a cheerful and positive tone.",
) as response:
    response.stream_to_file(speech_file_path)

Louis G 0

I found this project to perform the same actions as OpenAI

https://github.com/Azure-Samples/azure-openai-tts-demo

import asyncio
import os
from pathlib import Path
from openai import AzureOpenAI
from dotenv import load_dotenv
load_dotenv()
client = AzureOpenAI(
    azure_endpoint=os.getenv("AZURE_OPENAI_ENDPOINT"),
    api_key=os.getenv("AZURE_OPENAI_API_KEY"),
    api_version="2025-03-01-preview"
)
speech_file_path = Path(__file__).parent / "speech.mp3"
with client.audio.speech.with_streaming_response.create(
    model="gpt-4o-mini-tts",
    voice="coral",
    input="Today is a wonderful day to build something people love!",
    instructions="Speak in a cheerful and positive tone.",
) as response:
    response.stream_to_file(speech_file_path)

But I get the same error.

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

My model is deployed in eastus2 as was the github project.

Louis G 0 Reputation points

2025-10-08T16:39:02.9933333+00:00

Someone else has reported this issues as well.

https://github.com/openai/openai-python/issues/2418
SRILAKSHMI C 10,885 Reputation points Microsoft External Staff Moderator

2025-10-15T15:06:41.84+00:00
Hi Louis G,

I sincerely apologize for the delayed response.

You are correct at this time, the instructions parameter used in the OpenAI native API for the gpt-4o-mini-tts model is not yet supported in the Azure OpenAI API surface. This is why your API call returns the TypeError: Speech.create() got an unexpected keyword argument 'instructions', even though it works in the OpenAI platform’s playground and SDK examples.

To clarify:

Azure Playground UI may expose parameters like instructions that are available in the OpenAI API, but the corresponding Azure REST API version has not yet been updated to support them.

Currently, the Azure TTS capability for gpt-4o-mini-tts accepts the following key parameters: model, voice, input, and format but not instructions.

Style or tone guidance can instead be provided within the input text itself (e.g., “In a cheerful tone, say: ...”) as a temporary workaround until the feature parity update is released.

Example (Current Supported Approach in Azure):

with client.audio.speech.with_streaming_response.create( model="gpt-4o-mini-tts", voice="coral", input="Please speak in a cheerful and positive tone. Today is a wonderful day to build something people love!", ) as response: response.stream_to_file(speech_file_path)

Current Status:

The Azure API surface for gpt-4o-mini-tts is still in preview and not yet fully aligned with the latest OpenAI endpoints.

The instructions field and other enhancements are planned for inclusion in a future API version, but there is no confirmed ETA yet.

Please keep an eye on updates in the Azure OpenAI release notes and the Azure OpenAI SDK changelog for when this capability becomes available.

Thank you!
SRILAKSHMI C 10,885 Reputation points Microsoft External Staff Moderator

2025-10-16T16:50:07.21+00:00

Hi Louis G,

Did you get any chance to check the above response. Thank you!

Share via

Discrepancies between Azure Playground / OpenAI native API and Azure API

1 answer

Your answer