OpenAI: Whisper 1

OpenAI: Whisper 1 is an automatic speech recognition model from OpenAI, distributed through its API under the model id whisper-1. Whisper itself is OpenAI's open-source speech recognition family, and this endpoint exposes that capability as a hosted service. It sits in the transcription category, so its job is turning recorded speech into written text rather than generating conversational replies or reasoning over prompts.

The model takes audio as its only input and returns a transcription as its output. It handles both transcription and translation, and covers more than 50 languages. Audio files can be up to 25 MB in size, and it accepts a range of common audio formats. There is no text context window to manage here, because the model does not consume or produce chat-style token sequences; the unit of work is an audio file, and the result is the corresponding text. The model page does not list any tunable request parameters, so you work with the defaults the endpoint provides.

Because the output is text derived from speech, the practical uses follow directly. You can transcribe meeting recordings, interviews, lectures, podcasts or voice notes into searchable, editable text. The translation capability means audio in one of the supported languages can be rendered into English text, which is useful for multilingual archives or support recordings. Subtitle and caption generation for video is another natural fit, as is building a text layer over a library of call recordings for later review or indexing. Any workflow where audio needs to become written content, at file sizes up to 25 MB, is a candidate.

Calling it through api14 works the same way as any other model on the gateway, using the OpenAI-compatible API with the model id openai/whisper-1. You send your audio to the transcription endpoint in the format the OpenAI API expects, and the response contains the transcribed text. Since the supported parameters list is empty, there are no extra knobs documented for this model on api14; you pass the audio and the model id and read back the transcription. If your application already speaks the OpenAI API, switching to this model is a matter of changing the model identifier.

One thing to keep in mind is the 25 MB file limit. Longer recordings will need to be split before submission, and if your audio exceeds that size you will have to segment it yourself or preprocess it. Also note that this model only produces transcriptions; if you need a model that answers questions, follows instructions or generates text, a chat or language model is the better choice. Whisper 1 is the right pick when the task is strictly speech to text.