Skip to main content
POST
Text-to-Speech

GPT-Realtime 2.1 Mini

gpt-realtime-2.1-mini reads text into a complete audio file, including Russian speech. Check account availability in the catalog. Interactive voice sessions and user-facing streaming are planned for a later stage. opus produces an Ogg/Opus file. pcm is raw signed 16-bit audio at 24 kHz, mono; playback requires those parameters. MP3 and WAV are suitable for normal download and playback.

Price

The server derives the bucket from prompt. There is no additional debit after generation. Request a preliminary quote from POST /api/v2/models/gpt-realtime-2.1-mini/calculate-price, passing the text as params.full_text with the selected controls. Use the current quote for other TTS models.

Request

This asynchronous request returns HTTP 202 with a numeric id, status and tokens_spent. Poll GET /api/v2/generate/{id} until completed or failed. Download the result from GET /api/v2/download/{id} with the same Authorization header. If acknowledgement is lost, reconcile the existing job before submitting again; a new POST starts new work. After catalog cutover, openai-tts, openai-tts-hd and gpt-4o-mini-tts are unavailable for new selections. Historical results and prices remain. Legacy voices and speeds up to 4× are not copied to the successor automatically. For recognition, use the separate Speech-to-Text endpoint.