Audio Transcription

POST /v1/audio/transcriptions

Fully compatible with the OpenAI Audio Transcription API. Send audio as multipart/form-data; the OpenAI SDK does this automatically.

cURL

curl https://router-api.0g.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer sk-YOUR_API_KEY" \
  -F "file=@recording.mp3" \
  -F "model=openai/whisper-large-v3" \
  -F "response_format=json"

Python (OpenAI SDK)

from openai import OpenAI

client = OpenAI(
    base_url="https://router-api.0g.ai/v1",
    api_key="sk-YOUR_API_KEY",
)

with open("recording.mp3", "rb") as f:
    result = client.audio.transcriptions.create(
        model="openai/whisper-large-v3",
        file=f,
        response_format="json",
    )

print(result.text)

Fields

Field	Description
model	Audio model ID from /v1/models
file	Audio file (multipart form)
response_format	json, text, srt, verbose_json, vtt
language	ISO-639-1 code, e.g. "en" — optional, improves accuracy
prompt	Optional text to guide style and vocabulary
temperature	Sampling temperature (0 = deterministic)

Response

{
  "text": "Hello, this is a transcription of the audio file."
}

0G Router Extensions

Because this endpoint uses multipart/form-data instead of a JSON body, Router extensions are passed out-of-band — as a query parameter for verify_tee, and as X-0G-Provider-* request headers for provider routing.

TEE verification — pass verify_tee as a query parameter:

?verify_tee=true

See Verifiable Execution for what tee_verified means in the response.

Provider routing — pass X-0G-Provider-* headers, the same surface as JSON endpoints:

curl https://router-api.0g.ai/v1/audio/transcriptions \
  -H "Authorization: Bearer sk-YOUR_API_KEY" \
  -H "X-0G-Provider-Sort: latency" \
  -F "file=@recording.mp3" \
  -F "model=openai/whisper-large-v3"

See Provider Routing for the full header reference.

Related

• Models — list available audio models
• Verifiable Execution