GPT-4o Mini TTS
Steerable text-to-speech — tell it how to say it, not just what to say
About
GPT-4o Mini TTS launched on March 20, 2025 alongside OpenAI's new transcription models, replacing tts-1 as the recommended voice model. Built on the GPT-4o mini architecture, its headline feature is steerability: an instructions parameter lets developers control tone, emotion, accent, pacing, and style in plain language — "speak like a mad scientist" or "use a serene voice, like a mindfulness teacher" actually changes delivery. It ships with 11 to 13 preset voices (alloy, ash, ballad, coral, echo, fable, onyx, nova, sage, shimmer, verse, plus later additions), covers 100+ languages, and streams over SSE. Honest limits: voices are presets only — there is no voice cloning in this model (Custom Voice is a separate gated program); input caps at 2,000 tokens / 4,096 characters per request; there are no word-level timestamps; and non-English prosody is weaker than English. OpenAI positions it for voice agents, where the Realtime API now handles the conversational loop.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Team-reported: more natural and steerable than tts-1/tts-1-hd
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: OpenAI launch March 20, 2025 corroborated by TechTicia, ETCentric, OpenAI snapshot gpt-4o-mini-tts-2025-03-20; Steerable instructions param documented in OpenAI docs and third-party SDK references; Preset voices only, 4096-char limit, no word timestamps — honest limits captured.