OpenAI TTS Prompts

OpenAI's steerable text-to-speech voices — tell the voice how to speak in plain English, and it takes the direction.

audio

About OpenAI TTS

OpenAI TTS turns text into speech with a distinctive twist: the voices are steerable. Alongside the script you provide a plain-language instruction — how fast, how warm, how dramatic — and the voice adjusts its delivery to match. No markup language, no tag syntax: you direct it the way you would direct a voice actor.

That steerability is the whole point. The same script can come out as a calm meditation guide, a breathless sports commentator, or a sympathetic customer-service agent, purely by changing the instruction. And because it lives in the OpenAI API, it slots naturally into apps — narration pipelines, voice interfaces, and agents that speak.

The craft, then, is in writing the instruction, and it is less obvious than it looks: vague direction gets a default read, while a precise persona and pacing note transforms the delivery. The gallery below collects tested OpenAI TTS prompts — script plus instruction — with their real audio, so you can hear which phrasings actually steer.

How to write OpenAI TTS prompts

1Keep two layers cleanly separated: the script is what gets said, the instruction is how it gets said. Stage direction that leaks into the script gets read aloud verbatim.
2Steer with a persona, not just adjectives: "speak like a calm meditation guide leading a first session" outperforms "calm, soothing" because it implies pace, warmth, and pausing all at once.
3Spell out pacing and energy in plain words — "slow and deliberate, with generous pauses" or "brisk and upbeat, like sharing exciting news".
4Name the emotion and its intensity: "mildly amused" and "barely containing laughter" are different performances. One emotion per instruction keeps the read coherent.
5In the API, the instruction rides alongside the input text and combines with your chosen base voice — the same instruction lands differently per voice, so test a couple before committing.
6Iterate on the instruction, not the script: tweaking a single phrase of direction changes the delivery far more than rewording what is spoken.

Frequently asked questions

How do I change the tone of an OpenAI TTS voice?

Write the tone into the instruction in plain language — persona, mood, pacing, even attitude. The voice follows natural-language direction, so "warm, unhurried, like reassuring a friend" is a valid and effective control. The tested prompts on this page pair scripts with instructions that demonstrably work.

What's the difference between the script and the instruction?

The script is the exact text the voice will say; the instruction describes how to say it. Keeping them separate is the core skill — direction placed inside the script gets spoken aloud, while direction in the instruction shapes the performance invisibly.

Can I use OpenAI TTS in my own app?

Yes — it is designed for that. Through the OpenAI API you send the input text, a voice choice, and a steering instruction, and receive audio back. That makes it a natural fit for narration features, voice interfaces, and agents that need to speak with a specific character.

Where can I try these OpenAI TTS prompts?

OpenAI TTS is available through the OpenAI API, and OpenAI provides a demo playground for quickly auditioning voices and instructions. Copy a script-plus-instruction pair from this page and adjust the persona to fit your product.