Ordalin
Submit a tool

Reviewed tool profile

Muse Voice Transcribe

Meta's speech-to-text API model for live and recorded audio, with speaker labels, turn detection and 25 languages at $0.18 per hour.

Sources checked

Visit website ↗
Pricing
Pay-as-you-go at $0.18 per hour of audio processed ($3 per 1,000 minutes), metered in whole seconds. Streaming and file transcription cost the same, and there are no minimums or upfront commitment. Official source 1 ↗ · Official source 2 ↗
Available as
Meta Model API: a realtime WebSocket endpoint for live audio and a multipart HTTP endpoint for recordings, both running the same model. Official source 1 ↗
Primary group
Voice & Speech

Also consider

See all 6 similar tools ↓

What it does

Muse Voice Transcribe is Meta's speech-to-text model on Meta Model API, part of the Muse model family alongside Muse Spark and Muse Wonder. It is a developer API (model ID muse-voice-transcribe-1.0), not the voice chat in the consumer Muse app: it turns live audio streams or recorded files into text, with speech-turn detection, speaker labels for 20+ speakers, and keyword biasing across 25 languages. It returns text only and does not synthesize speech.

Official source 1 ↗ · Official source 2 ↗ · Official source 3 ↗

Features

  • Streaming transcription. Transcribes live audio over a WebSocket and sends partial transcripts that update while the speaker is still talking. Official source 1 ↗
  • Speech-turn detection. Endpointing mode lets the model find where each utterance ends and return one completed transcript per turn, without a separate voice-activity detector. Official source 1 ↗ · Official source 2 ↗
  • Speaker diarization. Diarization mode labels speakers such as A and B as the audio streams; Meta states attribution for 20+ speakers. Official source 1 ↗ · Official source 2 ↗
  • 25 languages with code-switching. Supports 25 languages, including English, Mandarin Chinese, Hindi, Spanish, Japanese and Arabic, and handles code-switching between them. Official source 1 ↗
  • Keyword biasing. A keyword list improves recognition of product names, people, places and acronyms without fine-tuning. Official source 1 ↗ · Official source 2 ↗
  • File transcription. A single HTTP request transcribes a WAV recording and returns JSON, server-sent events or plain text. Official source 1 ↗

Pricing & free limits

Billing details

Audio is metered in whole seconds and billed as its own line item. Official source 1 ↗

Zero data retention costs the same as Standard mode; the discounted Contributor mode is not offered for this model. Official source 1 ↗

Billing on Model API requires an admin to add a payment method and business details. Official source 1 ↗

Free access & limits

Free access and its limits were not established by the checked official sources.

Best for & limitations

Voice agents that need live partial transcripts and to know when a speaker has finished. Official source 1 ↗

Meeting and call analytics with speaker-labelled, turn-level transcripts. Official source 1 ↗

Live captions and dictation while audio is still arriving. Official source 1 ↗

Batch transcription of recordings you already have. Official source 1 ↗

  • No word-level timestamps. Only turn-level timestamps are returned; confidence scores, sound and emotion detection, and transcript reformatting are not available. Official source 1 ↗
  • File size and length caps. File uploads must be mono 16-bit WAV at 16 or 24 kHz, under 32 MB and 10 minutes per request. Official source 1 ↗
  • Speaker labels are per session. Labels such as A and B are not names and don't carry across sessions; diarization mode is not tuned for low-latency voice commands. Official source 1 ↗
  • Developer API only. Using it requires a Model API key and code; availability can vary by country. Official source 1 ↗ · Official source 2 ↗
Muse Voice Transcribe product screenshot
Website previewFixed desktop viewport · 1440 × 900

Similar tools

Published tools that share its categories.

TellPipA Mac call assistant that suggests what to say from your documents and client memory, with local transcription.
ProductivityFree trialChecked Nov 5
Visit
VowiseVoice dictation and transcription app that turns spoken notes into a journal, reviews and AI memory.
Voice & SpeechFree planChecked Nov 3
Visit
Audio8 ASREdge0's open speech-to-text models: live 24/7 Chinese/English streaming, plus a 0.1B model for CPUs and iPhones.
Voice & SpeechFreeChecked Oct 31
Visit
CluinnAI dictation for any Mac app, plus a call assistant that answers questions mid-call and lists action items after.
Voice & SpeechPaidChecked Oct 31
Visit