Sponsored by Byond Boundrys Consulting - Empowering Ideas, Delivering Results
OpenAI Intermediate

GPT-Live-Transcribe

GPT-Live-Transcribe is OpenAI’s low-latency streaming speech-to-text model for live captions, meetings and call analytics, billed at $0.017 per minute.

Speech & Voice ModelsTextAudio Paid
In plain English

What is this model and why does it matter?

GPT-Live-Transcribe is OpenAI's low-latency streaming speech-to-text model for live transcript deltas, with tunable latency, keyword hints, language hints and unstructured contextual guidance.

Live captionsMeeting transcriptionCall analyticsVoice agentsRealtime speech applications
Model overview

GPT-Live-Transcribe: features, use cases and important details

GPT-Live-Transcribe is a current AI model verified from first-party OpenAI sources.

GPT-Live-Transcribe verified specifications

GPT-Live-Transcribe is OpenAI’s low-latency streaming speech-to-text model for live transcript deltas, with tunable latency, keyword hints, language hints and unstructured contextual guidance. Its verified context or usage limit is Streaming realtime audio transcription workflow, with Transcript text deltas maximum output.

GPT-Live-Transcribe pricing and access

GPT-Live-Transcribe is billed by realtime audio duration at $0.017 per minute.

GPT-Live-Transcribe best uses

Live captions, Meeting transcription, Call analytics, Voice agents, Realtime speech applications.

GPT-Live-Transcribe limitations

Audio quality and network latency affect results, Not a general-purpose reasoning model, Requires realtime-session integration.

Get started

How to use this model

  1. Create an OpenAI API key.
  2. Open a realtime transcription session.
  3. Select gpt-live-transcribe.
  4. Stream live audio.
  5. Optionally provide keyword, context or language hints.
  6. Consume transcript deltas as they arrive.
Copy and try

Example prompts

  • Transcribe this live meeting with low latency.
  • Use these product names as keyword hints.
  • Detect and transcribe this multilingual customer call.
Capabilities

What it can do

  • Streaming speech-to-text
  • Low latency
  • Keyword hints
  • Language hints
  • Tunable latency
  • Transcript deltas
Best for

Practical use cases

  • Captions
  • Meetings
  • Contact centers
  • Voice assistants
  • Live accessibility
Pricing

What does it cost?

GPT-Live-Transcribe is billed by realtime audio duration at $0.017 per minute.

Input$0.017 / audio minute
OutputTranscript included in per-minute price
Simple summarySimple duration-based billing at $0.017 per live audio minute.

What stands out

  • Simple per-minute price
  • Purpose-built realtime transcription
  • Hint support
  • Low latency

Things to consider

  • No function calling
  • No structured outputs
  • No fine-tuning
Limitations

Important restrictions and trade-offs

  • Audio quality and network latency affect results
  • Not a general-purpose reasoning model
  • Requires realtime-session integration
SimplifyAITools verdict

Our editorial take

A practical current OpenAI speech model for developers building live captions, meetings and realtime voice-agent pipelines.

References

Primary sources

  1. Open source 1 ↗
  2. Open source 2 ↗
Keep exploring

Related AI models