Meta: Muse Voice Transcribe 1.0

Meta: Muse Voice Transcribe 1.0 is a speech-to-text model from Meta. It sits in the transcription category and is the only model in that family listed here under the Meta provider. Meta released it on 11 September 2026. If you are building anything that needs to turn spoken audio into written text, this is the model the provider positions for that job.

The model takes audio as its input and returns a transcription as its output. It is a synchronous model, which means it processes a request and returns the result within the same call rather than through an asynchronous job you poll later. Meta describes it as suited to push-to-talk, endpointing, and speaker-aware transcription. It also supports keyword biasing for domain terms and language biasing. The context length is not published in the facts available here, and the maximum output is listed as zero, so treat the output length as something to verify against your own workload rather than assume.

Those capabilities point to a fairly specific set of uses. Push-to-talk fits voice interfaces where a user holds a control, speaks, and expects text back immediately. Endpointing fits live audio streams where the model needs to decide when a speaker has finished an utterance, which matters for turn taking in voice assistants and meeting tools. Speaker-aware transcription fits recordings with more than one participant, where you want to know who said what. Keyword biasing is useful when your audio contains product names, internal jargon, or other terms a general model would mishear, and language biasing helps when you know the language in advance.

You call the model through api14's OpenAI-compatible API using the model id "meta/muse-voice-transcribe-1.0". Because the interface follows the OpenAI format, you can point an existing OpenAI client at the api14 base URL and send audio in the request rather than rewriting your integration around a provider-specific SDK. The facts list no supported parameters for this model, so do not assume you can pass temperature, top_p, or similar sampling controls. Check the live model page before relying on any optional field.

There are a couple of reasons to look elsewhere. The model only accepts audio, so if your input is text, images, or documents, it is not the right choice. It only produces transcriptions, so it will not summarise, translate, or answer questions about the audio content. If you need those behaviours, pair this model with a text model, or pick a model whose input and output modalities match your task.