gpt-4o-mini-transcribe - A lightweight speech-to-text model optimized for resource-constrained devices.

## Purpose of gpt-4o-mini-transcribe The **gpt-4o-mini-transcribe** model is designed for **lightweight speech-to-text transcription and translation**, optimized for **resource-constrained devices** such as mobile and embedded systems. It balances performance with efficiency, making it suitable for real-time applications like voice assistants, live transcription, and multilingual translation. ## Supported audio formats The model supports the following audio formats: - **mp3** - **mp4** - **mpeg** - **mpga** - **m4a** - **wav** - **webm** The maximum file size allowed is **25 MB**. ## Output Formats for Transcriptions The model provides two output formats: - **JSON** (structured data for developers) - **Text** (plain transcription output). ## Real-time streaming support Yes, **gpt-4o-mini-transcribe supports real-time streaming** for completed audio recordings. Developers can enable streaming using the `stream=True` parameter in API calls, receiving incremental transcription updates via `transcript.text.delta` events and a final `transcript.text.done` event. ## Handling unsupported languages For languages not explicitly supported (via ISO 639-1/639-3 codes), users can **specify the output language** using prompts (e.g., "Output in English"). The model will then transcribe and/or translate the audio accordingly. ## Pricing of the model As of March 2025, the model is priced at **$3 per million audio input tokens** (approximately **$0.003 per minute**), reflecting its cost-efficient design for scalable deployment. ## Integration method Developers can use the **OpenAI API endpoint (`/audio/transcriptions`)** with the Python SDK. Example code: ### Citation sources: - [gpt-4o-mini-transcribe](gpt-4o-mini-transcribe) - Official URL Updated: 2025-04-01