gpt-4o-mini-tts-2025-12-15

Overview. GPT-4o Mini TTS (snapshot 2025-12-15) is OpenAI's dedicated text-to-speech model built on the GPT-4o Mini foundation. It synthesizes spoken audio directly from text inputs via the dedicated speech API endpoint. Developers reach for this snapshot when they need predictable, standalone voice synthesis without the overhead of an end-to-end conversational pipeline.

Key strengths. The model supports thirteen built-in voices alongside custom voice profiles, rendering audio across formats including MP3, AAC, Opus, FLAC, WAV, and PCM. Its interactive latency profile supports responsive audio delivery for playback workflows. Specialized speech generation provides natural vocal inflection for simple narration tasks.

When to use it. Use this model for standard text-to-speech pipelines, audio notifications, and content readouts where input text is already finalized. Because it is optimized for simple, single-turn synthesis in the interactive latency class, it integrates well into lightweight media generation flows. For new production deployments, note that OpenAI has marked this snapshot deprecated with an end-of-life date of January 6, 2027, recommending migration to gpt-realtime-2.1-mini.

Limitations. The model operates strictly as a speech synthesis service and cannot perform text generation, conversational reasoning, or tool calling. Input size is restricted to a maximum of 2,000 tokens or 4,096 characters per request, requiring long-form documents to be pre-chunked. The snapshot is scheduled for API retirement in January 2027.

Sovereignty note. Operated exclusively on OpenAI infrastructure in the United States, this proprietary model is subject to the US CLOUD Act. It does not qualify as Sovereign and cannot be deployed in isolated European environments.