Music & audio

Music and audio-output models (Lyria 3, GPT Audio and others) don't have a dedicated endpoint: you call them through Chat completions and ask for audio output with modalities. The generated audio comes back base64-encoded in the assistant message.

POST https://api14.com/api/v1/chat/completions

Request

All chat completion parameters apply. The audio-specific ones:

Example

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api14.com/api/v1",
  apiKey: process.env.API14_API_KEY,
});

const result = await client.chat.completions.create({
  "model": "google/lyria-3",
  "messages": [
    {
      "role": "user",
      "content": "Explain vector databases in two sentences."
    }
  ]
});
console.log(result.choices[0].message.content);

A raw request:

{
  "model": "google/lyria-3",
  "modalities": ["text", "audio"],
  "messages": [
    { "role": "user", "content": "A 30-second lo-fi hip hop loop with soft piano and vinyl crackle." }
  ]
}

Response

{
  "id": "gen-1791100000-xyz",
  "object": "chat.completion",
  "model": "google/lyria-3",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Here is your loop.",
        "audio": { "id": "audio_01", "data": "UklGRiQAAABXQVZFZm10...", "format": "wav" }
      },
      "finish_reason": "stop"
    }
  ],
  "usage": { "prompt_tokens": 24, "completion_tokens": 1500, "total_tokens": 1524, "cost": 0.06 }
}

Decode choices[0].message.audio.data from base64 and save it with the given format extension.

Audio input

To send audio into a model (e.g. "summarize this recording"), use an input_audio content part, see Vision & files. For plain speech-to-text, Transcription is simpler and cheaper; for text-to-speech, use Speech.

Listing audio models

curl "https://api14.com/api/v1/models?category=audio"

Pricing

Audio models are billed at their listed price (per audio token, per second or per request, depending on the model). The exact cost is in usage.cost.

Errors

See Errors.