Model Overview
GPT-4o Mini TTS is a text-to-speech model based on GPT-4o mini that converts input text into natural speech. OpenAI released the model on March 20, 2025; the corresponding fixed version snapshot is gpt-4o-mini-tts-2025-03-20. The model alias and the fixed snapshot are not the same: OpenAI later pointed the alias to a newer snapshot. (openai.com)
Capabilities and Uses
- Text instructions can adjust aspects of speech delivery such as accent, emotion, intonation, speed, and tone, making it suitable for customer service voice applications, creative storytelling, and reading articles aloud. (openai.com)
- Supports multilingual speech generation, including Chinese, and streaming audio output. Voices are currently optimized for English. (developers.openai.com)
- Audio can be generated through the
v1/audio/speechendpoint, with support for formats including MP3, Opus, AAC, FLAC, WAV, and PCM. (developers.openai.com)
Limitations and Status
- The model accepts text input and outputs audio. It does not support image or video input and is not an audio-to-text model. The model page lists a maximum input of 2,000 tokens; the Speech API's
inputparameter has a separate limit of 4,096 characters. (developers.openai.com) - When used for end users, OpenAI requires clear disclosure that the voice they hear is AI-generated, not a human voice. (developers.openai.com)
- As of November 19, 2026, OpenAI had marked the model and its March 20, 2025 snapshot as deprecated, with shutdown scheduled for January 6, 2027. OpenAI recommends migrating new integrations to
gpt-realtime-2.1-miniin the Realtime API; migration requires changes to the API workflow. (developers.openai.com)
Pricing
As of November 19, 2026, the API prices listed on the OpenAI model page were in US dollars (USD): $0.60 per 1 million input text tokens; $12.00 per 1 million output audio tokens. These are the prices listed on the model page, not separate prices published specifically for the 2025 snapshot; separate pricing for that snapshot has not been verified. (developers.openai.com)