Meta、初のリアルタイム音声認識モデル「Muse Voice Transcribe」 20人超の話者識別と多言語混在に対応
ITmedia AI+ja
Metaは、リアルタイム音声認識モデル「Muse Voice Transcribe」を発表した。単一モデルで音声認識、話者分離、発話終了検知を処理し、20人以上の話者識別や日本語を含む多言語に対応する。「Meta Model API」で提供するほか、Mac版「Meta AI」」アプリや「Muse Code」でも利用可能だ。
This is a short summary published by AI Global Wire. The full article is owned and hosted by ITmedia AI+ — open it there to read it in full.
Read the full story at ITmedia AI+- Meta
Related AI news
- Asymmetries in Spontaneous and Instructed DeceptionarXiv cs.AI · September 2, 2026
- Manus resumes independent operations after Meta deal collapsesDIGITIMES · September 2, 2026
- AI 代理生態成關鍵考量,Meta 內部通訊工具棄 Google Chat 改用 SlackTechNews (TW) · September 2, 2026
- AI 代理生態成關鍵考量,Meta 通訊工具棄 Google Chat 改用 SlackTechNews (TW) · September 2, 2026
- Memo: Alexandr Wang says Meta is switching from Google Chat to Slack for internal communications as Slack is the "strongest platform available today for agents" (Business Insider)Techmeme · September 1, 2026
- Looking for evidence that Meta's AI investments are paying off? Here are 2 waysCNBC Technology · September 1, 2026