Audio & Speech

Today's Stories

1. 1

   Microsoft slashes speech AI price to 10 cents per hour

   Microsoft AI released MAI-Transcribe-2, a speech-recognition model it claims is faster, more accurate, and cheaper than offerings from OpenAI, Google, or ElevenLabs. The early-bird price is 10 cents per hour of audio, down from $0.36 an hour for the first model five months ago—a reduction of roughly 72%. For an enterprise processing 100,000 hours of call-center audio a year, the annual bill drops from $36,000 to $10,000. At this level, transcription becomes a negligible cost, potentially reshaping budgeting decisions for large banks and telecoms.

   This launch is part of Microsoft's strategy to build its own frontier-class models and swap them into products that once ran on OpenAI's technology. Transcription is the area where this plan has progressed fastest.

2. 2

   Vocaloid: The Soundtrack to Liminal Spaces

   The article explores how Vocaloid, Yamaha's singing-synthesizer software, has been used by Japanese artists to create uncanny, liminal moods. It highlights the music video for Tokyo Manaka's 'Brain Rot,' which features visuals inspired by 'the Backrooms' meme, and notes that the film 'Backrooms' has grossed over $394 million. Vocaloid, especially Hatsune Miku introduced in 2007 by Crypton Future Media, has become a cultural tool for expressing the unsettling aesthetics of liminal spaces and internet-era anxieties like social media addiction, predating the mainstream success of 'Backrooms.'

   The film 'Backrooms,' directed by 21-year-old Kane Parsons, arrives in Japanese theaters this week. Its viral success, alongside Vocaloid's influence, suggests growing mainstream interest in the liminal space aesthetic.

3. 3

   SoundHound AI stock predicted to rise 25% in a year

   SoundHound AI stock trades at about $7, near the bottom of its 52-week range, after swinging between less than $6 and more than $22 over the past 12 months. The author predicts the stock will rise by roughly 25% over the next year or so, though the path may remain volatile. SoundHound AI's voice AI platform lets people interact with cars and devices or do business without a human employee, and the company has expanded from autos into restaurants, hospitality, and customer service. Its pending purchase of LivePerson, at an enterprise value of about $250 million, adds a client base of 25 Fortune 100 companies, and management believes the combined company could generate between $350 million and $400 million in revenue in 2027.

   The author's upside scenario assumes 2027 revenue lands at $375 million—the middle of that guidance range—and values the stock at 12 times 2027 revenue, implying a market cap of about $4.5 billion and roughly 25% upside from a hypothetical $3.5 billion base value. However, SoundHound AI is not profitable, so investors should be careful about how high a valuation they bake in.

4. 4

   AI voice fraud costs ¥4.5B, ex-police warns

   A former Tokyo Metropolitan Police investigator with expertise in fraud cases warns that AI is being used in voice phishing scams that impersonate company executives, with reported damages reaching ¥4.5 billion. The advancement of generative AI is making it harder to distinguish real voices from fakes, putting business leaders at risk of being deceived through phone calls alone, without verifying identity through other means.

   The article highlights the latest voice phishing tools and fraud trends for 2026, offering guidance for organizations on how to respond to these evolving threats.

5. 5

   Voice AI brings hands-free ops to central kitchens

   Advanced Media and AWS Japan co-wrote this post about a solution shown at AWS Summit Japan 2026. It pairs the AmiVoice speech recognition engine with AI agents built on Amazon Bedrock AgentCore Runtime and the Strands Agents SDK. A serverless backend using Amazon DynamoDB and AWS Lambda completes the architecture for hands-free recipe operation. Central kitchen workers often have both hands occupied and must remove sanitary gloves to check recipes or log inventory, causing interruptions and hygiene risk. Manual record-keeping also leaves digital data uncollected, which limits management visibility; the solution records structured data automatically in the background as staff speak.

   AmiVoice also offers on-device and on-premises SDK options for environments where audio cannot leave the site, although AWS Marketplace currently lists only the cloud AmiVoice API. The post frames the pattern as applicable beyond food, with AI agents expected to perform tasks autonomously in next-generation workflows.

6. 6

   Meta Releases Muse Voice Transcribe, a Real-Time Speech Recognition Model

   Meta has announced Muse Voice Transcribe, its first real-time speech recognition model developed by Meta Superintelligence Labs (MSL). It is available from September 1 through the Meta Model API at $0.18 per hour. The model can distinguish more than 20 speakers and handle audio over an hour without post-processing, using a single model for streaming recognition, speaker separation, and end-pointing. It was trained on 70+ languages, with 25 verified at launch, including Japanese, and natively handles code-switching.

   Meta claims it ranks #1 in Artificial Analysis' streaming speech recognition and public speaker separation benchmarks. Its word error rate is 3.1% for finalized transcripts, with a 0.16-second delay from end of speech, beating Cartesia Ink-2 (3.4%) and ElevenLabs' Scribe v2 Realtime (3.6%). The model is also integrated into Mac's Meta AI and Muse Code for voice input.