Audio (Speech-to-Text)

AI Foundation Services provides Whisper-based audio models for transcription and translation, compatible with the OpenAI Audio API.

Prerequisites

• An API key (get one here)
• OpenAI SDK or HTTP client installed (Quickstart)
• An audio file (MP3, WAV, M4A, etc.)

What you'll learn:

• How to transcribe audio to text in the original language
• How to translate audio from any language to English
• Available audio models and parameters

---

List Audio Models

Audio models have model_type: "STT" in their metadata. Use the models endpoint and filter:

Python

from openai import OpenAI

client = OpenAI()

models = client.models.list()
for model in models.data:
    if model.meta_data.get("model_type") == "STT":
        print(model.id)

curl

curl "$OPENAI_BASE_URL/models" \
  -H "Authorization: Bearer $OPENAI_API_KEY"
# Filter results for models with model_type "STT"
# Available audio models: whisper-large-v3, whisper-large-v3-turbo

---

Audio Transcription

The transcription API converts audio into text in the same language as the input. It auto-detects the language from the first 30 seconds if language is not specified.

Python

from openai import OpenAI

client = OpenAI()

with open("/path/to/audio_file.mp3", "rb") as audio_file:
    transcription = client.audio.transcriptions.create(
        model="whisper-large-v3",
        file=audio_file,
        # language="en"  # Optional: specify language
    )

print(f"Transcription: {transcription.text}")

curl

curl -X POST "$OPENAI_BASE_URL/audio/transcriptions" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -F "model=whisper-large-v3" \
  -F "language=en" \
  -F "temperature=0.0" \
  -F "file=@/path/to/audio_file.mp3"

Example output:

Transcription: The stale smell of old beer lingers. It takes heat to bring out the odor.
A cold dip restores health and zest. A salt pickle tastes fine with ham.

---

Audio Translation

The translation API translates audio from any language into English.

Python

from openai import OpenAI

client = OpenAI()

with open("/path/to/audio_file.mp3", "rb") as audio_file:
    translation = client.audio.translations.create(
        model="whisper-large-v3",
        file=audio_file,
        temperature=1.0,
    )

print(f"Translation: {translation.text}")

curl

curl -X POST "$OPENAI_BASE_URL/audio/translations" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -F "model=whisper-large-v3" \
  -F "temperature=1.0" \
  -F "file=@/path/to/audio_file.mp3"

---

Parameters

Parameter	Type	Description
model	string	Audio model ID (e.g., whisper-large-v3)
file	file	The audio file to process
language	string	Optional. ISO language code. Auto-detected if omitted.
temperature	float	0.0 for deterministic, higher for more varied output

---

Key Features

1. Auto Language Detection — Identifies input language from the first 30 seconds
2. Customizable Output — Adjust behavior with language and temperature parameters
3. Efficient Processing — Low latency for both transcription and translation

---

Next Steps

• Chat Completions — Process transcribed text with LLMs
• Asynchronous Requests — Queue long audio files for async processing
• API Endpoints — Full endpoint reference