← Models

GPT-4o mini TTS

OpenAI · GPT-4o Audio

OpenAI text-to-speech model for responsive, API-first voice output workflows.

Part of GPT family · Other versions: GPT-5.3-Codex , GPT-5.3 , GPT-5.4-Pro , GPT-5.4 , GPT-5.2-Codex , GPT-5.2-Pro , GPT-5.2 , GPT-5-Codex , GPT-5 nano , GPT-5 mini , GPT-5 , GPT-4.1 , GPT-4o mini Transcribe , GPT-4o Transcribe , GPT-4o mini , GPT-4o
Type
audio
Context
N/A
Max Output
N/A
Status
current
API Access
Yes
License
proprietary
text-to-speech voice audio realtime api
Released March 2025 · Updated March 6, 2026

Overview

Freshness note: Model capabilities, limits, and pricing can change quickly. This profile is a point-in-time snapshot last verified on February 15, 2026.

GPT-4o mini TTS is OpenAI’s text-to-speech model tier for generating voice output in interactive applications and automation flows. It is intended for teams that want fast, tightly integrated TTS inside an existing OpenAI-centered stack.

Capabilities

The model supports programmatic voice generation for assistant responses, narrated content, and audio feedback loops. It is especially useful in systems already using OpenAI APIs for reasoning, orchestration, or realtime voice features.

Technical Details

For TTS, token context and max output fields are set to 0 in this content system and should be interpreted as N/A. Operational evaluation should prioritize voice quality, latency, and stability across languages and speaking styles.

Pricing & Access

OpenAI’s current pricing docs list GPT-4o mini TTS at $12.00 per 1M text input tokens. Because pricing, available voices, and voice controls can change by surface, confirm current details through official OpenAI documentation before launch.

Best Use Cases

Best for voice assistants, spoken notifications, educational narration, and multimodal interfaces needing low-friction speech output.

Comparisons

Compared with Eleven v3, GPT-4o mini TTS offers tighter OpenAI ecosystem integration but usually less emphasis on expressive voice performance. Compared with Realtime voice pipelines, it can be simpler for non-live or semi-live generation flows. Product-specific listening tests should drive final selection.