gpt-4o-mini-tts

Overview. GPT-4o Mini TTS is OpenAI's dedicated text-to-speech model derived from the GPT-4o Mini family. Accessed through the speech audio endpoint rather than chat completions, it converts plain text inputs of up to 2,000 tokens directly into synthesized speech audio. OpenAI has marked the model as deprecated, with an end-of-life shutdown scheduled for January 6, 2027.

Key strengths. The model provides 13 built-in voices and exports audio across multiple formats, including MP3, AAC, FLAC, Opus, WAV, and PCM. Operating within an interactive latency profile, it delivers dependable speech synthesis designed for user notifications, application prompts, and compact spoken passages.

When to use it. Deploy this model for straightforward text-to-speech pipelines that synthesize passages within the 2,000-token limit without requiring two-way voice conversations. Because OpenAI has deprecated the endpoint in favor of gpt-realtime-2.1-mini, teams initiating new implementations should plan for migration ahead of the January 2027 retirement date.

Limitations. Input size is capped at 2,000 tokens per request, requiring manual text chunking for longer documents. The model does not accept audio inputs, cannot return text completions or structured data, and supports no tool calling or cognitive reasoning capabilities.

Sovereignty note. GPT-4o Mini TTS is a proprietary closed-weight model operated by OpenAI in the United States. Inference requests fall under US jurisdiction and are subject to the CLOUD Act, without sovereign European hosting options.