Text to Speech API: gpt-4o-mini-tts & tts-1-hd

Choose a TTS model

Model	Best for	Input limit	Verified price
`gpt-4o-mini-tts`	Controllable delivery with `instructions`	4,096 characters; 2,000 model tokens	$0.60 / 1M text tokens + $12 / 1M audio tokens
`tts-1-hd`	Legacy HD TTS compatibility	4,096 characters	$30 / 1M characters

Audio created by these models is AI-generated. Tell end users that the voice is synthetic when you publish or play it.

gpt-4o-mini-tts

Use instructions to describe tone, pace, accent or emotion. The request below returns MP3 bytes; --output prevents curl from printing binary data to your terminal.

cURL · controlled MP3

curl https://api-models.com/v1/audio/speech \
  -H "Authorization: Bearer $API_MODELS_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o-mini-tts",
    "input": "Your order is ready for pickup.",
    "voice": "marin",
    "instructions": "Speak warmly, clearly, and at a relaxed pace.",
    "response_format": "mp3",
    "speed": 1.0
  }' \
  --output speech.mp3

Python · requests

import os
import requests

response = requests.post(
    "https://api-models.com/v1/audio/speech",
    headers={"Authorization": f"Bearer {os.environ['API_MODELS_KEY']}"},
    json={
        "model": "gpt-4o-mini-tts",
        "input": "Your order is ready for pickup.",
        "voice": "marin",
        "instructions": "Speak warmly, clearly, and at a relaxed pace.",
        "response_format": "mp3",
    },
    timeout=120,
)
response.raise_for_status()
with open("speech.mp3", "wb") as output:
    output.write(response.content)

Node.js · fetch

import { writeFile } from "node:fs/promises";

const response = await fetch("https://api-models.com/v1/audio/speech", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.API_MODELS_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "gpt-4o-mini-tts",
    input: "Your order is ready for pickup.",
    voice: "marin",
    instructions: "Speak warmly, clearly, and at a relaxed pace.",
    response_format: "mp3",
  }),
});
if (!response.ok) throw new Error(await response.text());
await writeFile("speech.mp3", Buffer.from(await response.arrayBuffer()));

tts-1-hd

Use the same endpoint without instructions. Compatibility voices alloy, echo, fable, onyx, nova and shimmer are the safe choice for this upstream; nova and alloy were exercised in live requests.

cURL · WAV output

curl https://api-models.com/v1/audio/speech \
  -H "Authorization: Bearer $API_MODELS_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tts-1-hd",
    "input": "A high-quality compatibility voice sample.",
    "voice": "nova",
    "response_format": "wav",
    "speed": 0.9
  }' \
  --output speech.wav

Parameters and output formats

Field	Requirement	Notes
`model`	Required	`gpt-4o-mini-tts` or `tts-1-hd`
`input`	Required	Maximum 4,096 characters
`voice`	Required	Use a voice compatible with the selected model
`instructions`	gpt-4o-mini-tts only	Controls delivery; not supported by tts-1-hd
`response_format`	Optional	`mp3`, `opus`, `aac`, `flac`, `wav`, `pcm`; default `mp3`
`speed`	Optional	0.25 to 4.0; default 1.0

The gateway currently documents normal binary responses only. Do not add stream_format=sse; it is not part of the verified compatibility surface.

Common errors

Symptom	Fix
Chat completion error	Use `/v1/audio/speech`, not `/v1/chat/completions`
Empty or corrupt file	Save the binary response and check `response.ok` before writing
Unsupported voice behavior	Use the compatibility voices listed for `tts-1-hd`
instructions rejected	Remove it from `tts-1-hd` requests