aiaimodel.cards

GPT-4o mini TTS

Audio modelPrevious generationCard summarized

Text-to-speech model powered by GPT-4o Mini

● Re-checked 12 Oct 2026 · 49 facts · 1 correction · 1 unconfirmedSource: "Updates for developers building with voice" (OpenAI Developers blog, Dec 2025) and text-to-speech guide; no system card published

Use it via API

$12.00 / 1M tok
Code exampleCompare

Where to call it

  • OpenAI APIgpt-4o-mini-tts

Supports

  • Streaming
Full feature support

Good for

  • Narrating written content
  • Spoken audio in multiple languages
  • Realtime audio output via streaming

Watch out for

Overview

OpenAI · 02mini, front

4o mini TTS

Text-to-speech model powered by GPT-4o Mini

Typical price
$12.00 / 1M tok
Generated output per 1M tokens
Max length
—
Languages
Follows Whisper's ~57 languages; voices optimized for English
Modes
text-to-speech, streaming
Released
20 Mar 2025

OpenAI published no system card for GPT-4o mini TTS. It is a text-to-speech model built on GPT-4o mini that can be steered with instructions for accent, emotion, intonation, speed and tone. The 2025-12-15 snapshot, the default since 13 Jan 2026, has about 35% lower word error rate on Common Voice and FLEURS than the snapshot it replaced. OpenAI deprecated it on 1 Oct 2026, and it is removed from the API on 6 Jan 2027. model page; TTS guide; blog; changelog; deprecations page

gpt-4o-mini-tts
Typical price
$12.00 / 1M tok
Languages
Follows Whisper's ~57 languages; voices optimized for English
Weights
Closed
Released
20 Mar 2025
  • text-to-speech
  • streaming
  • steerable delivery (instructions)
  • custom voices (eligible customers)