Keep pulling the thread on Google DeepMind.
Google DeepMind's VO3 model can natively generate audio simultaneously with video from a single prompt.
Olivia observes that Apple appears to be outsourcing many of its advanced AI features to OpenAI's ChatGPT, which will run on Apple devices, rather than building them in-house.
Eleven Labs' 11 V3 model allows users to control vocal emotions, inflections, and accents through text-based "tags" in the prompt.
Based on A16Z data, the median consumer AI startup now reaches an Annual Recurring Revenue (ARR) of $4.2 million within 12 months of monetizing.
According to A16Z data, the top quartile of consumer AI startups reach $8.7 million in ARR and the bottom quartile reach $2.9 million in ARR within 12 months of monetizing.
A16Z's data indicates that consumer AI startups are ramping revenue approximately twice as fast as B2B AI startups.
According to A16Z data, consumer AI startups charge an average of $22 per month, more than double the average subscription price for consumer companies in the pre-AI era.
Justine asserts that the Flux Context image model maintains character and object consistency across edits significantly better than OpenAI's GPT-4o image model.
Olivia believes Google's VO3 model represents a "ChatGPT moment" for AI video, signifying a major breakthrough in quality and public awareness.
Google's VO3 video model is currently limited to generating clips that are a maximum of eight seconds long.
The VO3 model only generates audio for video clips created from a text prompt, not for those generated from an image-to-video prompt.
Access to Google's VO3 model via the Flow creative studio initially required a subscription to the $250 per month Google AI Ultra plan.