Voice instructions and presets
Control tone, pace and style of OpenAI voices in Text to Speech using natural-language instructions and eight ready-made presets.
Written By Umakhan Magomedov
Last updated 4 days ago
When using OpenAI voices in Text to Speech, you can provide a natural-language instruction that tells the AI how the voice should sound. You can write your own instruction or choose one of the built-in presets.
ℹ️ Instructions only work with OpenAI voices (Alloy, Shimmer, Echo and others). They are not supported for MiniMax voices or Custom Voices.
How to set an instruction
Open Text to Speech and select an OpenAI voice.
Scroll down to the Instructions section.
Tap the text field and type a description of how the voice should sound.
Or tap Presets and choose one of the ready-made styles.
Tap Generate. The instruction is sent along with your text and influences the output.
Available presets
Writing custom instructions
You can write any instruction in plain language. The model understands descriptions of tone, pace, emotion and style. Examples:
"Speak with a slight British accent. Slow pace. Thoughtful pauses."
"Sound like a friendly customer service agent. Warm and helpful."
"Energetic sports commentator voice. Fast pace, excitement."
ℹ️ Instructions are a prompt to the AI, not a guaranteed transformation. Results may vary between voices and languages. If the output does not match, try rephrasing the instruction.