OpenAI Releases GPT Live Transcribe for Streaming Audio
OpenAI released GPT Live Transcribe on July 28, 2026. This model supports low-latency streaming transcription with keyword hints and multi-language input via…
Release Details and Scope
OpenAI officially released GPT Live Transcribe on July 28, 2026. This announcement coincided with the release of GPT Transcribe. The source identifies GPT Live Transcribe as a model designed for low-latency streaming transcription. It is distinct from GPT Transcribe, which handles accurate file transcription and final transcripts of committed Realtime turns.
- GPT Live Transcribe targets low-latency streaming scenarios.
- GPT Transcribe targets file transcription and final Realtime turn transcripts.
- The release date recorded in the official changelog is July 28, 2026.
Supported Features and Inputs
Both GPT Live Transcribe and GPT Transcribe support specific input configurations. The documentation states that both models accept free-form transcription context. They also support keyword hints to guide the transcription process. Additionally, both models handle multiple expected input languages. The input modality for GPT Live Transcribe is audio, while the output modality is text.
- Supports free-form transcription context.
- Accepts keyword hints.
- Handles multiple expected input languages.
- Input modality: Audio; Output modality: Text.
Evaluation Procedure and Limitations
Engineers can evaluate GPT Live Transcribe by testing it against specific audio streams. A practical procedure involves measuring latency during real-time sessions and comparing the accuracy of keyword hint application against baseline runs without hints. Users should validate the model’s handling of multiple expected input languages by providing mixed-language audio samples. The source confirms that open weights are not confirmed for this model. It does not establish whether local installation is possible or if API access is mandatory. No pricing data, benchmark scores, or performance rankings are provided in the source material.
- Test streaming audio via the v1/realtime endpoint.
- Measure response latency under load.
- Compare transcription accuracy with and without keyword hints.
- Validate multilingual input handling with diverse audio samples.
Primary sources
Frequently asked questions
What is the difference between GPT Live Transcribe and GPT Transcribe?
GPT Live Transcribe is designed for low-latency streaming transcription. GPT Transcribe is designed for accurate file transcription and generating final transcripts of committed Realtime turns. Both share support for context, keyword hints, and multiple languages.
Can I run GPT Live Transcribe locally?
The source states that open weights are not confirmed. It does not provide information regarding local installation requirements or availability. Therefore, the ability to run the model locally remains unverified based on the provided facts.