Try MiniMax FastH3 from Fastvideo →
Sign InStart Building

Task History

Click to see output list

No tasks yet

Go to Models
Audio & SpeechActive

openai / gpt-realtime-mini

gpt-realtime-mini

byopenai

GPT Mini Realtime enables low-latency, bidirectional streaming for voice and text. Build interactive, responsive AI experiences that feel natural and immediate.

Speech to SpeechRealtime ConversationFast InferenceVoice Agent
Model ID
gpt-realtime-mini
Provider
openai
Updated
1787043327
wiro playground—openai/gpt-realtime-mini
Reset to defaults

Required. The voice of the AI assistant.

Required. Instructions that define the AI assistant's behavior and personality.

Sample outputs
Sample 1
Updated 1787043327

GPT Realtime Voice Assistant

Overview

A real-time voice assistant built using OpenAI's GPT models, designed for natural conversation and voice interaction. This tool enables seamless voice communication with AI assistants.

What you can build

  • Voice-enabled chatbots
  • Interactive voice assistants
  • Real-time voice-to-voice conversational systems
  • Audio-based AI application interfaces

Inputs

  • Voice Selection: Choose from 13 different voices for the AI assistant
  • System Instructions: Define the AI assistant's behavior and personality
  • Audio Formats: Configure input/output audio formats (PCM, G.711 μ-law, G.711 A-law)
  • Sample Rates: Set input/output audio sample rates (24kHz recommended)
  • Transcription Model: Select transcription model (GPT-4o, Whisper 1, etc.)
  • Turn Detection Settings: Adjust sensitivity and silence duration thresholds

Outputs

  • Natural voice responses in selected audio format
  • Transcribed user speech when using transcription models
  • Real-time conversational feedback

Recommended settings

  • Use 'Marin' or 'Cedar' voices for best naturalness
  • Set audio sample rate to 24000Hz for optimal quality
  • Use 'GPT-4o Transcribe' for high-quality transcription
  • Set turn detection threshold to 0.5 for balanced responsiveness

Limitations

  • Performance depends on internet connection quality
  • Audio processing may introduce slight latency
  • Voice recognition accuracy varies by audio environment
  • Not optimized for extremely noisy environments

Safety & compliance

  • Designed for general conversational use
  • Audio data processing complies with applicable privacy regulations
  • No personal data retention beyond session scope
  • Intended for authorized users only

API quick start

Run gpt-realtime-mini with a single API call.

POST https://api.wiro.ai/v1/Run/openai/gpt-realtime-mini
{
  "voice": "marin",
  "system_instructions": "You are a helpful voice assistant. Speak …",
  "transcription_model": "gpt-4o-mini-transcribe",
  "input_audio_format": "audio/pcm"
}
View full API docs