Skip to main content

​
Overview

AudioPod exposes OpenAI-compatible audio endpoints, so code written for the OpenAI audio API runs against AudioPod with two changes only: the base URL and the API key. Use the official OpenAI SDKs — no AudioPod-specific client required.
EndpointPurpose
POST /v1/audio/speechText to speech — returns audio bytes
POST /v1/audio/transcriptionsSpeech to text (source language)
POST /v1/audio/translationsSpeech to text, translated to English
These three are the only audio shapes the OpenAI API defines. AudioPod’s other capabilities — stem separation, music generation, speaker separation, voice cloning, noise reduction, media conversion — are available through the native REST API, SDKs and MCP server.

​
Configuration

Set the base URL to https://api.audiopod.ai/api/v1 and use your AudioPod API key (starts with ap_). Create one in your dashboard. 
from openai import OpenAI

client = OpenAI(
    base_url="https://api.audiopod.ai/api/v1",
    api_key="ap_your_api_key",
)
Calls are billed to your API wallet (prepaid, in USD). Nothing is charged for a request that fails — you’re only billed once audio or a transcript is produced.

​
Text to speech

POST /v1/audio/speech synthesizes input with the requested voice and returns the audio bytes. 
response = client.audio.speech.create(
    model="tts-1",
    voice="nova",
    input="Hello from AudioPod.",
    response_format="mp3",  # mp3 | opus | aac | flac | wav | pcm
)
response.stream_to_file("speech.mp3")
FieldNotes
inputThe text to synthesize (required).
voiceA voice name (e.g. nova, onyx) or a voice ID. List options at GET /api/v1/voice/voice-profiles.
response_formatmp3 (default), opus, aac, flac, wav, pcm.
speed0.25–4.0 (default 1.0).

​
Transcription

POST /v1/audio/transcriptions converts speech to text in the source language. 
with open("audio.mp3", "rb") as f:
    transcript = client.audio.transcriptions.create(
        model="whisper-1",
        file=f,
        response_format="json",  # json | text | verbose_json | srt | vtt
    )
print(transcript.text)
FieldNotes
fileThe audio file to transcribe (required).
languageOptional ISO-639-1 hint (e.g. en); auto-detected when omitted.
promptOptional text to guide the transcription.
response_formatjson (default), text, verbose_json, srt, vtt.
verbose_json includes the detected language, duration, and segment-level timestamps.

​
Translation

POST /v1/audio/translations converts speech in any language to English text. The request shape matches transcription (the language field is not used — output is always English). 
with open("spanish.mp3", "rb") as f:
    translation = client.audio.translations.create(
        model="whisper-1",
        file=f,
    )
print(translation.text)  # English

​
Notes & differences

  • Auth: pass your AudioPod API key as a Bearer token (the scheme the OpenAI SDKs use). X-API-Key is also accepted.
  • Models: the model field is accepted for compatibility. AudioPod selects the appropriate engine for each request; you don’t need to change model strings when migrating.
  • Voices: common OpenAI voice names map to AudioPod voices. Browse the full catalog — including custom voice clones — at GET /api/v1/voice/voice-profiles and pass any voice name or ID as voice.
  • Errors follow the OpenAI shape ({ "error": { "message", "type", "param" } }), so existing OpenAI error handling works unchanged.

​
Beyond OpenAI

Already migrated? AudioPod does more than the OpenAI audio API. Explore stem separation, music generation, speaker separation, and voice cloning through the native API and SDKs.