Chat GPT's voice clone AI app is too dangerous to release. OpenAI's Speech Engine is an AI model that generates speech that mimics the speaker's voice from a 15-second audio sample. It can be used for purposes such as educational support, translation, and support for people who don't speak the language. What could go wrong? Because of the potential for misuse, OpenAI is taking a cautious approach to adoption, allowing only a small number of companies to test new tools. Why test this if it's too risky to publish? What I'm really concerned about is that if Open AI can do this, so can others. And they don't issue press releases. At Schwab, my voice is my password. Gulp.
Higgsfield AI launches Diffuse. The company is led by former Snap head of AI Alex Mashrabov, who sold his previous startup AI Factory to Snap for $166 million. The Diffuse mobile app can generate videos from text prompts and use selfies to generate clips starring that person. It's currently free, but the company plans to charge social media content creators and marketers a fee to create short videos. Sora has also been tested by media professionals, but it is computationally expensive and will definitely be priced for businesses rather than individuals. Higgfield's generative model was developed by a team of 16 people in less than nine months. Funding for Higgfield comes from his $8 million seed round led by Menlo Ventures.
Bitkraft Ventures raises $275 million for third interactive fund. This is the venture capital firm's third fund. Bitkraft Venture Fund 3 invests in studios, platforms, and technologies in the gaming and interactive media categories at seed and Series A stages. Their portfolio includes Frost Giant, Anzu, Carry1st, InWorld, Voicemod, Immutable, and Karate Combat. Bitkraft runs his six venture funds with a portfolio of 130 companies.
Meta's Reality Labs celebrates 10 years of Oculus VR acquisition. Since then, Meta has spent more than $10 billion on XR efforts. It's no exaggeration to say that it currently costs more than $100 billion. CNet's XR editors took a trip back in time with Andrew Bosworth, the chief technology officer who oversees the effort. Bos says that while display in regular glasses remains a challenge, adding AI to wearables like smart glasses will open up new ways of thinking about how XR will evolve. .
Spatial Persona for Apple's Vision Pro. Technically, Persona has been available since the headset launched on February 2nd, but it received a major upgrade this week that has everyone talking. Personas are created by scanning your face with a headset, but previously they existed within the frame. Persona only has a head, shoulders, and hands, but facial tracking captures facial expressions, creating a sense of realism. As the person you're talking to moves around, the headset transforms your perspective as if you were there.
This column, formerly called “This Week in XR,” is also a podcast It will be hosted by author Charlie Fink, former Red Camera studio executive and co-founder Ted Shirowicz, and Magic Leap founder Ronnie Abowitz. This week's guest is Maureen Fan, the CEO of Baobab Studios. You can find us on Spotify, iTunes, and YouTube.
what we are reading
AI could actually end the world (Sherry Palmer's blog)
Sora's best 7 videos (Tom's Guide)