翻訳待ち：Meta just beat OpenAI and Google at real-time transcription

AI サービスが一時的に利用できないため、復旧後に翻訳を補完します。

Meta’s Superintelligence Labs on Tuesday launched Muse Voice Transcribe, a new real-time speech recognition model that, at least on some benchmarks, outperforms virtually every other comparable model when it comes to working with speech in real time. Meta’s lab describes the model as its first “real-time audio perception model.” With Muse Spark, the company also recently shipped another speech-to-text capable model, though not one that specializes in this use case. The model can distinguish between more than 20 speakers, Meta says, and has been trained on over 70 languages (with 25 of them having been “extensively verified”), including when multi-lingual speakers switch languages in the middle of a conversation. It also supports long conversations of over an hour. Credit: Meta. It’s now available via the Meta Model API, Meta AI for Mac, and in Muse Code. The API pricing seems reasonable, at $3.00 per 1,000 audio-minutes (or $0.18 per hour). Sadly, unlike with its Muse Glimmer models, it doesn’t look like Meta will make the open weights of this model available. Benchmark lead, in English On Artificial Analysis’s AA-WER Streaming speech-to-text accuracy benchmark, the model’s word error rate is 3.1%, ahead of competitors like Cartesia Ink-2 (3.4%), ElevenLabs’ Scribe v2 Realtime (3.6%), GPT Live Transcribe (3.9%), and Gemini 3.5 Transcribe Live (4%). This benchmark only applies to speech in English, though. When it comes to recognizing distinct speakers, all models still struggle more than most users would like, but here, too, for these real-time use cases, Muse Voice Transcribe leads the pack with a 17.5% error rate across a number of standard benchmarks. Credit: Meta. How it all works Under the hood, Muse Voice Transcribe is an autoregressive multimodal model from the Muse Spark family, Meta says, and the interesting part is how it decides when to talk. Audio comes in as 80-millisecond chunks (12.5 per second), each compressed into a single soft token. At every chunk, the model makes a choice. It either emits a text token or it emits a special “next audio” placeholder, which the system then swaps out for the next chunk of audio. When the audio stops, an “empty audio” token tells the model there’s nothing more coming, and it flushes whatever text it’s still holding. Because the model controls how much audio it hears before committing to a word, it also controls its own latency. Meta calls this “adaptive delay.” The idea here is that difficult words get more context, while easy words get transcribed almost immediately. That tradeoff is learned within the models’ reinforcement learning phase, where a word error rate reward and a delay reward are multiplied together rather than added. The system uses a similar mechanism for detecting speakers. Real-time transcription has quietly become one of the most crowded corners of the AI market this summer, with OpenAI, Google, xAI, and Alibaba all shipping streaming models within weeks of each other, on top of the specialists that were already there. A 0.3 point lead on a benchmark won’t hold for long in a competitive field like that. What Meta has, however, is a built-in reason to keep pushing. Every product it really cares about, from its glasses to the Mac app, needs this to work as well as possible. The post Meta just beat OpenAI and Google at real-time transcription appeared first on The New Stack.